Professional Documents
Culture Documents
Book Catalyst Design Optimal Distribution of Catalyst in Pellets Reactors and Membranes
Book Catalyst Design Optimal Distribution of Catalyst in Pellets Reactors and Membranes
Series Editor:
Arvind Varma, University of Notre Dame
Editorial Board:
Alexis T. Bell, University of California, Berkeley
John Bridgwater, University of Cambridge
Robert A. Brown, MIT
L. Gary Leal, University of California, Santa Barbara
Massimo Morbidelli, ETH, Zurich
Stanley I. Sandler, University of Delaware
Michael L. Shuler, Cornell University
Arthur W. Westerberg, Carnegie Mellon University
Massimo Morbidelli
ETH, Zurich
Asterios Gavriilidis
University College London
Arvind Varma
University of Notre Dame
To our teachers and students
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521660594
A catalogue record for this publication is available from the British Library
1 Introduction 1
1.1 Importance of Catalysis 1
1.2 Nonuniform Catalyst Distributions 1
1.3 Overview of Book Contents 4
2 Optimization of the Catalyst Distribution in a Single Pellet 6
2.1 The Case of a Single Reaction 6
2.1.1 Isothermal Conditions 6
2.1.2 Nonisothermal Conditions 15
2.1.3 Arbitrary Kinetics with External Transport Resistances 18
2.1.4 Dynamic Behavior 23
2.2 Multiple Reactions 25
2.2.1 Isothermal Conditions 25
2.2.2 Nonisothermal Conditions 28
2.3 The General Case of a Complex Reaction System 30
2.3.1 An Illustrative Example 31
2.3.2 General Reaction System 35
2.4 Catalyst Dispersion Considerations 40
2.4.1 Factors Affecting Catalyst Dispersion 40
2.4.2 Dependence of Catalytic Surface Area on Catalyst Loading 43
2.5 Optimal Distribution of Catalyst Loading 46
2.5.1 The Problem Formulation 47
2.5.2 A Single First-Order Isothermal Reaction 51
2.5.3 Linear Dependence between the Active Element Surface
Area and Its Loading 54
2.5.4 First-Order Nonisothermal Reactions: Numerical Optimization 55
2.5.5 Multistep Optimal Loading Distribution 59
2.6 Experimental Studies 63
2.6.1 Oxidation Reactions 63
2.6.2 Hydrogenation Reactions 66
2.6.3 Fischer–Tropsch Synthesis 68
vii
viii Contents
xi
xii Preface
composite zeolite, biological, and functionalized polymer resin types, are ad-
dressed in Chapter 6. The final Chapter 7 considers catalyst preparation by im-
pregnation techniques, where the effects of adsorption, diffusion and drying on
obtaining desired nonuniform catalyst distributions within supports are discussed.
This book should appeal to all those who are interested in design, preparation and
utilization of supported catalysts, including chemical and environmental engineers
and chemists. It should also provide a rich source of interesting mathematical prob-
lems for applied mathematicians. Finally, we hope that industrial practitioners will
find the concepts and results described in this book to be useful for their work.
This book can be used either as text for a senior-graduate level specialized
course, or as a supplementary text for existing courses in reaction engineering,
industrial chemistry or applied mathematics. It can also be used as a reference for
industrial applications.
We thank our departmental colleagues for maintaining an atmosphere con-
ducive to learning. We also thank our families for their encouragement and sup-
port, which made this writing possible.
Massimo Morbidelli
Asterios Gavriilidis
Arvind Varma
1
Introduction
1
2 Introduction
fixed-bed reactor. This is a parallel–consecutive reaction system for which the in-
termediate product yield is benefited by a pellet with an inert core. Verykios et al.
(1983) modeled ethylene epoxidation in a nonisothermal nonadiabatic fixed-bed
reactor with nonuniform catalysts. They showed that improved reactor stability
against runaway could be obtained, along with higher reactor selectivity and yield,
as compared to uniform catalysts.
Rutkin and Petersen (1979) and Ardiles (1986) studied the effect of activity
distributions for bifunctional catalysts in fixed-bed reactors, for the case of multiple
reaction schemes. Each reaction was assumed to require only one type of catalyst.
It was shown that catalyst activity distributions had a strong influence on reactant
conversion and product selectivities.
Nonuniform activity distribution for catalysts experiencing deactivation has
been studied by a number of investigators (DeLancey, 1973; Shadman-Yazdi and
Petersen, 1972; Corbett and Luss, 1974; Becker and Wei, 1977b; Juang and Weng,
1983; Hegedus and McCabe, 1984). If deactivation occurs by sintering, it is mini-
mized by decreasing the local catalyst concentration, i.e., a uniform catalyst offers
the best resistance to sintering (Komiyama and Muraki, 1990).
In all cases considered above, catalyst performance was assessed utilizing ap-
propriate indexes. The most common ones include effectiveness, selectivity, yield,
and lifetime. Effectiveness factor relates primarily to the reactant conversion that
can be achieved by a certain amount of catalyst, while selectivity and yield relate
to the production of the desired species in multiple reaction systems. In the case of
membrane reactors additional performance indexes (e.g. product purity) become
of interest. In deactivating systems, other indexes incorporating the deactivation
rate can be utilized apart from catalyst lifetime. Another index, which has not
been employed in optimization studies because it is difficult to express in quanti-
tative terms, is attrition. Catalyst pellets with an outer protective layer of support
are beneficial in applications where attrition due to abrasion or vibration occurs,
since only the inert and inexpensive support is worn off and the precious active
materials are retained.
The key parameters which control the effect of nonuniform distribution on the
above performance indexes are reaction kinetics, transport properties, operating
variables, deactivation mechanism, and catalyst cost. All the early studies dis-
cussed above demonstrated that nonuniform catalysts can offer superior conver-
sion, selectivity, durability, and thermal sensitivity characteristics to those wherein
the activity is uniform. This was done by comparing the performance of catalysts
with selected types of activity profiles, which led to the best profile within the
class considered, but not to the optimal one. Morbidelli et al. (1982) first showed
that under the constraint of a fixed total amount of active material, the optimal
catalyst distribution is an appropriately chosen Dirac-delta function; i.e., all the
active catalyst should be located at a specific position within the pellet. This dis-
tribution remains optimal even for the most general case of an arbitrary number
of reactions with arbitrary kinetics, occurring in a nonisothermal pellet with finite
external heat and mass transfer resistances (Wu et al., 1990a).
It is worth noting that optimization of the catalyst activity distribution is car-
ried out assuming that the support has a certain pore structure and hence specific
4 Introduction
effective diffusivities for the various components. Thus for a given pore structure,
the catalyst distribution within the support is optimized. An alternative optimiza-
tion in catalyst design is that of pore structure, while maintaining a uniform catalyst
distribution. In this case, the mass transport characteristics of the pellet are op-
timized. This approach has been followed by various investigators and has been
shown to lead to improvements in catalyst performance (cf. Hegedus, 1980; Pereira
et al., 1988; Hegedus and Pereira, 1990; Beeckman and Hegedus, 1991; Keil and
Rieckmann, 1994).
Much effort has also been invested in the preparation of nonuniformly active
catalysts. As insight is gained into the phenomena related to catalyst prepara-
tion, scientists are able to prepare specific nonuniform profiles. In this regard, it
should be recognized that catalyst loading and catalyst activity distributions are in
principle different characteristics. In catalyst preparation, the variable that is usu-
ally controlled is the local catalyst loading. However, under reaction conditions,
the local reaction rate constant is proportional to catalyst activity. The relation
between catalyst activity and catalyst loading is not always straightforward. For
structure-sensitive reactions, it depends on the particular reaction system, and
hence generalizations cannot be made. On the other hand, for structure-insensitive
reactions, catalyst activity is proportional to catalyst surface area. Thus, if the lat-
ter depends linearly on catalyst loading, then the catalyst activity and loading
distributions are equivalent. If the above dependence is not linear, then the two
distributions can be quite different. The majority of studies on nonuniform cat-
alyst distributions address catalyst activity optimization, although a few investi-
gators have considered catalyst loading optimization by postulating some type of
surface area–catalyst loading dependence (Cervello et al., 1977; Juang et al., 1981).
Along these lines, it was shown that when the relation between catalyst activity
and loading is linear, and the latter is constrained by an upper bound, the optimal
Dirac-delta distribution becomes a step distribution. However, if this dependence
is not linear, which physically means that larger catalyst crystallites are produced
with increased loading, then the optimal catalyst distribution is no longer a step,
but rather a more disperse distribution (Baratti et al., 1993). An important point
is that in order to make meaningful comparisons among various distributions, the
total amount of catalyst must be kept constant.
Work in the areas of design, performance, and preparation of nonuniform cat-
alysts has been reviewed by various investigators (Lee and Aris 1985; Komiyama
1985; Dougherty and Verykios 1987; Vayenas and Verykios, 1989; Komiyama and
Muraki, 1990; Gavriilidis et al., 1993a). In this monograph, these issues are dis-
cussed with emphasis placed on optimally distributed nonuniform catalysts. Spe-
cial attention is given to applications involving reactions of industrial importance.
extended to the most general case of an arbitrary number of reactions with arbi-
trary kinetics under nonisothermal conditions, in the presence of external trans-
port limitations. The analysis includes the effect of catalyst dispersion varying
with catalyst loading. Finally, the improved performance of nonuniform catalysts
is demonstrated through experimental studies for oxidation, hydrogenation, and
Fischer–Tropsch synthesis reactions.
Optimization of catalyst distribution in pellets constituting a fixed-bed reactor
requires one to take into account changes in fluid-phase composition and tem-
perature along the reactor. This is discussed in Chapter 3, for single and multiple
reactions, under isothermal and nonisothermal conditions. The discussion of ex-
perimental work is focused on catalytic oxidations.
Catalyst distribution influences the performance of systems undergoing deacti-
vation, and this issue is addressed in Chapter 4 for selective as well as nonselective
poisoning. Experimental work on methanation, hydrogenation, and NO reduction
is presented to demonstrate the advantages of nonuniform catalyst distributions.
In Chapter 5, the effect of catalyst distribution on the performance of inorganic
membrane reactors is discussed. In such systems, the catalyst can be located either
in pellets packed inside a membrane (IMRCF) or in the membrane itself (CMR).
Experimental results for an IMRCF are presented, and the preparation of CMRs
with controlled catalyst distribution by sequential slip casting is introduced.
In Chapter 6, special topics of particular industrial importance are discussed.
These include automotive catalysts, where various concepts of nonuniform dis-
tributions have been utilized; hydrotreating catalysts, which is a particular type
of deactivating system; composite catalysts, with more than one catalytic function
finding applications in refinery processes; biocatalysts; and functionalized polymer
resins, which find applications in acid catalysis.
The final Chapter 7 considers issues related to catalyst preparation. The discus-
sion is focused on impregnation methods, since they represent the most mature
technique for preparation of nonuniform catalysts. During pellet impregnation,
adsorption and diffusion of the various components within the support are im-
portant, and can be manipulated to give rise to desired nonuniform distributions.
The chapter concludes with studies where experimental results are compared with
model calculations.
2
Optimization of the Catalyst
Distribution in a Single Pellet
A
mong various reaction systems, investigation of optimal catalyst distribution
in a single pellet has received the most attention. Although the general prob-
lem of an arbitrary number of reactions following arbitrary kinetics occur-
ring in a nonisothermal pellet has been solved and will be discussed later in this
chapter, it is instructive to first consider simpler cases and proceed gradually to
the more complex ones. This allows one to understand the underlying physic-
ochemical principles, without complex mathematical details. Thus, we first treat
single reactions, under isothermal and nonisothermal conditions, and then analyze
multiple reactions.
6
2.1 The Case of a Single Reaction 7
is best, while for large values of φ (i.e. diffusion control) the outer is best. For
intermediate values of the Thiele modulus, the middle distribution has the highest
effectiveness factor. So the question naturally arises: given a Thiele modulus φ,
among all possible catalyst distributions, which one is the best? This question can
be answered precisely, and is addressed next.
A → products (2.1)
so that by definition
1
a(x) dVp = 1. (2.4)
Vp Vp
The constraint of a fixed total amount of catalyst means that k̄Vp is constant. In
dimensionless form, the above equations become
1 d n du
s = φ 2 a(s) f (u) (2.6)
s n ds ds
du
s = 0: =0 (2.7a)
ds
s = 1: u=1 (2.7b)
1
1
a(s)s n ds = (2.8)
0 n+1
where the following dimensionless quantities have been introduced:
Since we are dealing with a single reaction, the catalyst performance is directly
related to the effectiveness factor, defined by
1
f (u)a(s)s n ds
η = 01 (2.10)
a(s)s n ds
0
so that
r (C) (1 + σ )2 u
f (u) = = (2.13)
r (Cf ) (1 + σ u)2
where
σ = KCf . (2.14)
The shape of the rate function f (u) depends on the parameter σ and is shown in
Figure 2.2. In particular, f (u) has a unique maximum at
um = 1/σ. (2.15)
The dimensionless reaction rate reaches its maximum value in the range 0 < u < 1
for σ > 1, and at u = 1 for σ ≤ 1. Thus, summarizing, the Langmuir–Hinshelwood
kinetics exhibits a maximum value M at u = um , where
um = 1/σ, M = (1 + σ )2 /4σ for σ >1 (2.16a)
um = 1, M=1 for σ ≤1 (2.16b)
Since f (u) ≤ M, from the expression for η given by equation (2.11) it is evident that
1 1
η = (n + 1) f (u)a(s)s n ds ≤ (n + 1)M a(s)s n ds (2.17)
0 0
Therefore, for any activity distribution a(s), the effectiveness factor can never be
greater than M. It is apparent that if a function a(s) exists for which η = M, this
will constitute the solution of the optimization problem.
10 Optimization of the Catalyst Distribution in a Single Pellet
it can be easily shown that the activity distribution (2.19) is indeed the optimal
one. For this, equation (2.19) is substituted into equation (2.11) to give
1
δ(s − s̄ opt ) n
ηopt = f (u) n s ds = f (um ) = M. (2.22)
0 s̄ opt
φ02
s̄ opt = for n=2 (2.23c)
φ02 + 12(σ − 1)
where φ0 is a “clean” Thiele modulus which does not include the adsorption pa-
rameter σ and is defined as
φ02 = k̄R 2 /De = (1 + σ )2 φ 2 . (2.24)
2.1 The Case of a Single Reaction 11
σ≤1 σ>1
s̄ opt 1 1 − a (n = 0)
exp(−) (n = 1)
1
(n = 2)
1+
(1 + σ )2
ηopt 1
4σ
a For > 1, one has s̄ opt = 0 and ηopt = f (ū), where ū is the
solution of 1 − ū − φ 2 f (ū) = 0. Here f (ū) is given by equation
(2.13).
(1 + σ )2
ηopt = . (2.25)
4σ
The effect of all the involved physicochemical parameters on s̄ opt can be expressed
through a single dimensionless parameter , which is used in Table 2.1 to summa-
rize the results obtained so far:
4(n + 1)(σ − 1)
= . (2.26)
φ02
From inspection of equation 2.23 it appears that while for the infinite cylinder
(n = 1) and the sphere (n = 2) we have 0 ≤ s̄ opt ≤ 1 for all values of φ0 ≥ 0 and
σ ≥ 1, for the infinite slab (n = 0) the value of s̄ opt can become negative. This is
physically unrealistic, and in this case the optimal catalyst distribution is s̄ opt = 0,
i.e., the catalyst must be concentrated at the pellet center (Morbidelli et al., 1982).
The resulting value of the effectiveness factor will in this case be smaller than
ηopt given by equation (2.25), but it is still the maximum obtainable for the given
values of φ0 and σ .
From equations (2.23), it is seen that the catalyst location depends on the physic-
ochemical parameters of the system, i.e., the Thiele modulus φ0 and the adsorption
constant σ . On increasing the Thiele modulus or decreasing the adsorption con-
stant, the optimal location of the active catalyst moves from the interior of the
pellet to the external surface. This is shown in Figure 2.3, where for the spherical
pellet the s̄ opt -vs-φ0 curves for various σ values are plotted. Increasing the Thiele
modulus (keeping the adsorption constant unchanged) leads to larger diffusional
resistances, and therefore moves the location where u = um closer to the pellet’s
external surface. Similarly, decreasing the adsorption constant (keeping the Thiele
modulus unchanged) causes the maximum of the reaction rate to occur at larger
u values, which again are encountered closer to the external surface.
12 Optimization of the Catalyst Distribution in a Single Pellet
The system of equations (2.6)–(2.7) for the step distribution can only be solved
numerically. Care must be exercised in finding all possible solutions, since multiple
solutions may exist.
In Figure 2.4, the effectiveness factor is plotted as a function of Thiele modulus
for given values of σ, s̄, and . Note that the maximum value of the effectiveness
factor ηm is attained in a region of φ where multiplicity is present. This ηm value
is shown as a function of the step-distribution half thickness in Figure 2.5. As the
step width decreases, the behavior of the step distribution approaches that of the
σ>1
If this is at s̄ < s̄ opt , then the effectiveness factor is lower than its maximum value
to an extent proportional to the difference s̄ − s̄ opt . However, if the location of
the active layer is s̄ > s̄ opt , then the effectiveness factor undergoes a dramatic de-
crease. The reason is related to the pellet’s multiplicity behavior. As illustrated in
Figure 2.4, a certain level of diffusional resistance is required to give the maximum
value of the effectiveness factor. If the actual diffusional resistance is larger, i.e.
φ > φm , the effectiveness factor on the upper branch undergoes a slight decrease,
while if it is smaller, i.e. φ < φm , it undergoes a dramatic decrease, since the upper
branch is no longer available. Thus, in preparing catalysts, if an error is anticipated,
it is better to err towards placing the step deeper within the pellet than s̄ opt .
where
φ02
Bi∗ = . (2.32)
4(n + 1)(σ − 1)
A → products. (2.33)
The steady-state mass and energy balances within the pellet under nonisothermal
conditions are
1 d dC
De n xn = a(x)r (C, T) (2.34)
x dx dx
1 d dT
λe n xn = −(−H) a(x) r (C, T) (2.35)
x dx dx
dC dT
x = 0: = 0, =0 (2.36a)
dx dx
x = R: C = Cf, T = Tf . (2.36b)
The catalyst distribution a(x) for nonisothermal systems is defined as the ratio
between the local rate constant and its volume-average value, both evaluated at
the bulk fluid temperature:
σ ≤ σc σ > σc
s̄ opt 1 1 − a n=0
exp(− ) n = 1
1
n=2
1 +
ηopt 1 f (um ),
where um is given by f (um ) = 0
a For > 1, one has s̄ opt = 0 and ηopt = f (ū), where ū is the solu-
tion of 1 − ū − φ 2 f (ū) = 0. Here f (ū) is given by equation (2.45),
and σc is given by equation (2.47).
1 − βγ
σc = . (2.47)
1 + β(γ + 2ε)
The effects of all physicochemical variables are described through a single dimen-
sionless parameter
(n + 1)(1 − um )
= (2.48)
φ 2 f (um )
where um is the value of u which maximizes the reaction rate f (u) given by equa-
tion (2.45).
In general, a larger dimensionless heat of reaction β leads to higher tempera-
tures within the pellet, thus increasing the ratio between reaction and diffusion
rates. This makes the reactant concentration profile steeper and therefore makes
the optimal location approach the external surface of the pellet (see Figure 2.7).
However, for small values of σ the optimal location can exhibit a minimum as
a function of β, as illustrated in Figure 2.7. As β increases, the optimal location
moves from the surface (at β = 0) towards the pellet interior, reaches a minimum
value, and then moves back to the pellet surface (β → ∞).
18 Optimization of the Catalyst Distribution in a Single Pellet
βγ > 1. (2.49)
This behavior is solely due to the intrapellet temperature gradient, and indicates
that the temperature increase at the location of the reaction causes the net reaction
rate to increase sufficiently so as to more than overcome the adverse effect of its
decrease due to decreased reactant concentration. Due to the above temperature
gradient, effectiveness factors larger than unity can be attained.
θ = 1 + βµ + βus (1 − µ) − βu (2.53)
where µ = Bim /Bih and the subscript s refers to the value at the pellet external
surface, i.e. s = 1. Using equation (2.53), the system of two differential equations
with variables u and θ reduces to one differential equation
1 d n du
s = φ 2 a(s)d(u, us ) (2.54)
s n ds ds
where d(u, us ) is given by f (u, θ) with θ replaced by (2.53). Note that two variables
u and us are present, since the value of us is not known a priori. For this reason, the
procedure used in previous cases (sections 2.1.1 and 2.1.2) cannot be applied here.
Details of the original analysis by Chemburkar et al. (1987) are not presented here,
because the simpler and more general proof in section 2.3.2 can also be applied
to this case.
Let us now simply report the final results. The optimal catalyst location is given
by
where
1 uδ − um
e = · s . (2.56)
Bim 1 − uδs
Note that e is always positive, so that equations (2.55) always provide physically
meaningful values for s̄ opt , i.e. 0 < s̄ opt < 1. The only exception is for the slab
20 Optimization of the Catalyst Distribution in a Single Pellet
geometry (n = 0), where e can be larger than 1, in which case s̄ opt = 0. In order
to use the relation above we need to compute um and uδs . The first, um , is the value
of u which maximizes the function d(u, uδs ), while uδs is the smallest root of the
equation
(n + 1)Bim
1 − uδs = d um , uδs . (2.57)
φ2
This also provides the value of the optimal effectiveness factor, given by
ηopt = d um , uδs . (2.58)
The number of parameters upon which the active layer location depends is
now larger, and for the case of a nonisothermal mth-order reaction these are
the dimensionless heat of reaction β, the dimensionless activation energy γ , the
reaction order m, the geometry integer n, the mass Biot number Bim , the ratio
µ of mass to heat Biot numbers, and the Thiele modulus φ. These dimensionless
parameters in turn contain various physicochemical variables. However, given
a reaction, a fixed amount of catalyst, and a support, only two of them can be
considered as operating variables, viz. Tf and Cf . Any change in them affects
φ, β and γ , although the effect is not substantial for the latter two. Hence, the
dependence of the optimal location s̄ opt on the Thiele modulus φ is shown in
Figure 2.8 for typical values of the other parameters while one of them is varied
around its base value. The optimal catalyst activity location s̄ opt can exhibit a
discontinuity as a specific physicochemical parameter is varied, and this arises
from the fact that often multiple steady states are possible.
The effect of external transport resistances is illustrated in Figure 2.8.a and
b. As expected, when the resistances increase, i.e. when the mass or heat Biot
number decreases, the optimal location moves outwards. Also, the discontinuity
in the s̄ opt -vs-φ plot occurs at lower values of φ as the transport resistances in-
crease. The effect of pellet geometry (Figure 2.8.c) is that as n increases, the jump
occurs at a higher value of φ. It can also be seen in Figure 2.8.d–f that in general
as β increases (or γ increases, or m decreases): (1) the optimal location moves
inward for values of φ sufficiently small that no jump has occurred, and (2) the
jump occurs at smaller φ. However, for sufficiently small values of β or γ (or
sufficiently large values of m), the optimal location s̄ opt is always 1, regardless of
the value of the Thiele modulus. This behavior is directly related to the fact that
when
βγ
<1 (2.59)
m
for finite Bim and Bih , the optimal location for all φ is s̄ opt = 1 [compare with
equation (2.49)].
The discontinuity that the optimal catalyst location exhibits as a physicochemi-
cal parameter is varied results in a similar discontinuity in the effectiveness factor
2.1 The Case of a Single Reaction 21
Figure 2.8. Optimal location of catalyst s̄opt as a function of Thiele modulus φ for an mth-
order reaction. Bim = 50, µ = 10, n = 2, β = 0.05, γ = 25, and m = 1, unless otherwise
specified. (From Chemburkar et al., 1987.)
22 Optimization of the Catalyst Distribution in a Single Pellet
(see Figure 2.9). When the jump occurs from a lower to a higher s̄ opt value, the
higher value is typically 1, i.e. at the external surface of the pellet (see Figure
2.8.e), and the corresponding effectiveness factor also jumps by some orders of
magnitude (see Figure 2.9.a). So, if there is free choice of the operating variables
(particularly the bulk fluid values Tf and Cf ), then the optimal catalyst location
can always be moved to the external surface of the pellet. In addition to the
convenience of preparing eggshell catalysts, such a strategy will also result in an
increased catalytic effectiveness.
A1 + ν2 A2 → products (2.60)
does not affect the amplitude of oscillations significantly, for a step distribution
the amplitude increases as the Thiele modulus decreases. Furthermore, while for
the Dirac distribution the relation of steady-state effectiveness factor to φ 2 ln s̄
is independent of catalyst location s̄ (see Figure 2.11), for step-type catalysts it
depends on both step location and width.
Parallel reactions: A1 → A2 , A1 → A3
For parallel reactions that follow power-law kinetics the optimal location of the
Dirac-delta is at the surface of the pellet if the order of the desired reaction is
higher than the order of the undesired one; in the opposite case it is at the center
of the pellet. Egg-white distributions can be optimal for more complex kinetics.
As an example, in Figure 2.13 we consider the case where the desired reaction
follows bimolecular Langmuir–Hinshelwood kinetics, while the undesired reac-
tion follows unimolecular Langmuir–Hinshelwood kinetics, and the two involve
different adsorption sites. The dependence of the optimal catalyst location, se-
lectivity, and dimensionless reaction rate of the desired product on the ratio of
Thiele moduli, κ = (φ2 /φ1 )2 , is shown for the optimal Dirac, uniform, and exter-
nal shell catalyst distributions. The performance index used for optimization was
selectivity. When the undesired reaction becomes faster (i.e., κ increases), the se-
lectivities for all three types of distributions decrease. However, the Dirac-type
catalyst always gives the highest selectivity value. Simultaneously, the location of
the Dirac distribution moves towards the external surface of the pellet. The net
rate of production for the desired product is also consistently the highest for the
Dirac distribution. It is worth noting that the sensitivity of selectivity to catalyst
location behaves qualitatively in the same manner as discussed previously for the
effectiveness factor in the case of a single reaction.
Figure 2.13. Effect of the ratio of Thiele moduli, κ, on the optimal catalyst
location, s̄opt (solid lines), global selectivity S (dashed lines), and dimen-
sionless rate of production of the desired product A2 , f¯ (dotted lines),
for the optimal distribution (Sopt , f¯opt ), uniform distribution for slab ge-
ometry (Sun , f¯un ), and external shell distribution (Sext, f¯ext ). Parallel reac-
1 2
tions: A1 −→ A2 , A1 −→ A3 ; f1 (u1 ) = (1 + σ1 )2 u1 /(1 + σ1 u1 )2 , f2 (u1 ) =
(1 + σ2 )u1 /(1 + σ2 u1 ), σ1 = 5, σ2 = 20, φ1 [(n + 1)/gv ]1/2 = 1, De,1 /De,2 = 1.
Optimization is based on selectivity. (From Vayenas and Pavlou, 1987b.)
2.2 Multiple Reactions 27
Consecutive reactions: A1 → A2 → A3
When two consecutive reactions occur, where the desired one follows bimolecular
Langmuir–Hinshelwood kinetics and the undesired one follows linear kinetics, the
Dirac catalyst distribution again gives the highest selectivity for all values of κ,
as illustrated in Figure 2.14. The difference in selectivity obtained with different
distributions (Sopt , Sun , and Sext ) increases with κ. The optimal distribution not only
gives the highest selectivity, which is the optimization index, but also the highest
net rate of production of the desired species. However, it should be noted that the
optimal catalyst location for selectivity maximization is in general different from
the optimal catalyst location for maximum production of the desired product.
For the reaction system considered, the catalyst location is rather insensitive to
changes in κ, except for sufficiently high values of κ, where the optimal location
moves rapidly towards the center of the pellet, because diffusional resistances
slow down the increasingly faster first-order undesired reaction. Although a Dirac-
delta function is always the optimal distribution, the actual performance of Dirac
catalysts will depend strongly on the type of kinetics that the reactions follow, and
can be quite different than the specific cases discussed above.
Figure 2.14. Effect of the ratio of Thiele moduli, κ, on the optimal catalyst
location s̄opt (solid lines), global selectivity, S (dashed lines), and dimen-
sionless rate of production of the desired product A2 , f¯ (dotted lines), for
the optimal distribution (Sopt , f¯opt ), uniform distribution for slab geometry
(Sun , f¯un ), and external shell distribution (Sext , f¯ext ). Consecutive reactions:
1 2
A1 −→ A2 −→ A3 ; f1 (u1 ) = (1 + σ )2 u1 /(1 + σ u1 )2 , f2 (u2 ) = u2 , σ = 8,
φ1 [(n + 1)/gv ]1/2 = 1, De,1 /De,2 = 1, uf,2 = 1. Optimization is based on
selectivity. (From Vayenas and Pavlou, 1987b.)
28 Optimization of the Catalyst Distribution in a Single Pellet
Parallel reactions: A1 → A2 , A1 → A3
As an example consider two parallel reactions, where the desired one is exother-
mic, has low activation energy, and exhibits second-order kinetics, while the unde-
sired one is endothermic, has high activation energy, and exhibits linear kinetics
(Vayenas and Pavlou, 1988). The effect of the ratio of Thiele moduli, κ = (φ2 /φ1 )2 ,
on the optimal location and selectivity, where the latter is optimized, is shown in
Figure 2.15. In order to obtain a physical understanding of this system, it should be
Consecutive reactions: A1 → A2 → A3
The case of two consecutive exothermic reactions with linear kinetics and higher
activation energy for the first desired reaction is illustrated in Figure 2.16 (Vayenas
et al., 1989). Since the reaction orders are the same for both reactions, in this case
it is clear that concentration gradients do not affect selectivity at all. Increased
temperature favors the desired reaction, since its activation energy is larger than
that of the undesired one, i.e. γ1 > γ2 . Such temperatures are obtained by placing
the catalyst inside the pellet. For high values of κ, which indicates that the unde-
sired reaction is so fast that it approaches diffusion control, the optimal catalyst
location moves rapidly towards the center of the pellet.
A system of multiple reactions of great industrial importance, the ethylene
epoxidation reaction network, was investigated by Morbidelli et al. (1984a). Ethy-
lene reacts over silver catalyst to give ethylene oxide, which is the desired product,
and carbon dioxide, which is the undesired by-product. Ethylene oxide is also com-
busted, but this reaction is less important than the combustion of ethylene, and so
it was not included in the above study. It was shown that the selectivity to ethylene
oxide is maximized when the catalyst is located at the external surface of the pel-
let. This is expected for parallel exothermic reactions with similar kinetics when
the undesired reaction has a higher activation energy than the desired one. The
maximum selectivity is attained at the surface of the pellet, even if combustion
of ethylene oxide is included (Pavlou and Vayenas, 1990a). However, if the net
production rate of ethylene oxide is to be maximized, then subsurface locations
are optimal, a fact which arises from the nonisothermality of the pellet.
Numerical optimization
A numerical optimization technique has been developed by Baratti et al. (1990),
and applied to the consecutive–parallel reaction network
A1 → A2 → A3
(2.62)
A1 → A4 .
The technique is based on the method of orthogonal collocation over finite el-
ements, whereby the differential equations are reduced to a set of nonlinear
algebraic equations, whose solution is obtained through the Newton–Raphson
method. It was shown that the number of collocation points used in the calcula-
tions is critical, and if it is too small, then erroneous results can arise. In Figure
2.17, the numerically calculated optimal activity distribution is shown for the re-
acting system noted above, where all the reactions are exothermic and follow
first-order kinetics. The optimization index is the pellet effectiveness factor, and
it corresponds to maximum consumption rate of component A1 . The shape of the
obtained distribution closely resembles a Dirac delta function, in full agreement
with analytical calculations. This numerical technique will also be used in the opti-
mization of complex reacting systems, and will be discussed further in section 2.5.4.
The optimization problem consists in finding the function a(s) which maximizes
the performance index η under the constraint given by equations (2.63), (2.64),
and (2.66).
where both a(s) and a ∗ (s) satisfy the constraint (2.66). The dimensionless con-
centration profile u(s), resulting from a(s), can also be represented as a per-
turbation about the optimal concentration profile u∗ (s) resulting from a ∗ (s) as
follows:
where
d f (u)
with BCs
dλ
s = 0: =0 (2.77a)
ds
s = 1: λ = 0. (2.77b)
The last term on the right-hand side of equation (2.75), when integrated twice by
parts, becomes
1 1
λL[δu]s ds =
n
δu L[λ]s n ds (2.78)
0 0
Substituting the above relation into equation (2.75) and using equations (2.76)
and (2.77) leads to
1
δη = (n + 1) G ∗ δa s n ds (2.79)
0
where
G ∗ = (1 − φ 2 λ) f ∗
(2.80)
for arbitrary small ε > 0. The above statement is proven by contradiction, i.e. by
showing that if it is not true, and a ∗ (s) is substantially nonzero at least at one point
s = s̄ where
For this purpose, let us define a new distribution a(s), which is a small pertur-
bation of a ∗ (s), satisfying the constraint (2.66):
∗ n ∗
a (s) + γ [s + s − s̄] a (s + s − s̄)/s for s̄ − ε < s < s̄ + ε
n
Since γ > 0 is arbitrarily small, a(s) can be made infinitesimally close to a ∗ (s),
as required by equation (2.68). Therefore the necessary condition for optimality,
(2.81), can be used. From equation (2.84) it follows that
1 s̄+ε
G ∗ (s)a(s)s n ds = γ [s + s − s̄]n a ∗ (s + s − s̄)G ∗ (s) ds
0 s̄−ε
+ a ∗ (s)G ∗ (s)s n ds
(0,1)\(s −ε,s +ε)
s +ε
+ (1 − γ ) a ∗ (s)G ∗ (s)s n ds. (2.85)
s −ε
From equation (2.83) and the continuity of G ∗ (s), it follows that for sufficiently
small ε > 0
Using this equation, we determine a lower bound on the first integral on the right-
hand side of equation (2.85). For this, it is convenient to introduce the new variable
z = s + s − s̄, and then from equation (2.86) we obtain
s+ε
¯ s +ε
n ∗ ∗
[s + s − s̄] a (s + s − s̄)G (s) ds > G ∗ (s)a ∗ (s)s n ds. (2.87)
s̄−ε s −ε
which violates the necessary condition (2.81) for optimality. Since the distribution
a(s), which is a small perturbation about a ∗ (s), violates (2.81), a ∗ (s) cannot be
optimal. Therefore, in order for a ∗ (s) to be optimal, it has to be substantially
nonzero only at the point s̄ where the corresponding G ∗ (s) is maximum, which
means that a ∗ (s) must be a Dirac-delta function. Although it is highly unlikely that
G ∗ (s) will attain its maximum simultaneously at more than one point, it should
be mentioned that in this case the optimal distribution would take the form of
multiple Dirac-delta functions (Pavlou et al., 1991; Morbidelli et al., 1991; see also
section 2.5.5).
It is worth noting that the above proof fails when the optimal activity distribu-
tion a ∗ (s) generates a function G ∗ (s) which is constant and equal to its maximum
2.3 The General Case of a Complex Reaction System 35
The selectivity of the Mth component (reactant) towards the Nth component
(product) is defined as the ratio between the rate of production of component N
and the rate of consumption of component M:
1 J n
j=1 ν Nj φ j f j (u, θ ) s ds
2
0 a(s)
SNM = 1 J . (2.99)
j=1 ν Mj φ j f j (u, θ ) s ds
2 n
0 a(s)
The yield of desired product N with respect to reactant M is defined as the ratio
between the actual rate of production of component N and the rate of consump-
tion of reactant M which would have prevailed in the absence of all transport
2.3 The General Case of a Complex Reaction System 37
resistances:
1 J n
(n + 1) 0 a(s) j=1 ν Nj φ j f j s ds
2
YNM = η M SNM = J (2.100)
j=1 ν Mj φ j
2
with BCs
dx
s = 0: =0 (2.104a)
ds
dx
s = 1: = A(xf − x). (2.104b)
ds
The expressions for the performance indexes [equations (2.98)–(2.100)] take the
form
1
(n + 1) 0 a(s)w M (x)s n ds
ηM = (2.105)
DM Jj=1 ν Mj φ 2j
1
DM 0 a(s)w N (x)s n ds
SNM = (2.106)
DN 1 a(s)w M (x)s n ds
0
1
(n + 1) a(s)w N (x)s n ds
YNM = η M SNM = J
0
. (2.107)
DN j=1 ν Mj φ 2j
38 Optimization of the Catalyst Distribution in a Single Pellet
Since the expressions for the effectiveness factor (2.105) and yield (2.107) have
the same form, the objective function in these cases, up to a constant multiplier,
can be written in general as
1
U1 = (n + 1) a(s)wi (x)s n ds. (2.108)
0
The necessary conditions for optimality of effectiveness factor and yield are
1 1
∗ ∗
G1 a (s)s ds ≥
n
G1∗ a(s)s n ds (2.113)
0 0
We may observe that, even though the expressions for the G function are different
for the two optimization problems, the form of the optimality condition is the same.
The method utilized in the case of a single isothermal reaction to prove that the
optimal catalyst distribution is a Dirac-delta function is based on the existence
and form of the G function. Since the optimality conditions (2.113) and (2.114)
have the same form as (2.81), and furthermore, no matter how complicated the G
functions may be, when a ∗ is given, G1∗ and G2∗ are functions only of s, the same
arguments can be followed also in this case.
Thus, it can be concluded that, for any catalyst performance index (i.e. effec-
tiveness, selectivity, or yield) and for the most general case of an arbitrary number
of reactions, following arbitrary kinetics, occurring in a nonisothermal pellet, with
finite external mass and heat transfer resistances, the optimal catalyst activity distri-
bution is a Dirac-delta function.
2.3 The General Case of a Complex Reaction System 39
and the yield (2.100) towards the desired product N with respect to reactant M to
J
j=1 ν Nj φ j f j (ū, θ̄ )
2
YNM = η M SNM = J . (2.118)
j=1 ν Mj φ j
2
For a given catalyst location s̄, equations (2.119) and (2.120) are solved for ū and θ̄
and then the performance index of interest is calculated from equations (2.116)–
(2.118). Caution should be exercised during the numerical solution, because the
40 Optimization of the Catalyst Distribution in a Single Pellet
nonlinearity of the equations implies the possibility of multiple solutions. The op-
timal catalyst location s̄ opt is obtained by identifying the s̄ value in the interval
[0, 1] which maximizes the desired performance index. Note that for a given re-
acting system, the optimal catalyst location s̄ opt depends strongly on whether the
performance index considered is effectiveness, selectivity, or yield.
Impregnation
During impregnation, the active agent is introduced into the support as a so-
lution of a precursor compound, which is typically not its final form. Catalyst
dispersion is affected in this step by the interaction of precursor and support (e.g.
strong or weak adsorption). This in turn depends on the precursor properties and
concentration, nature of the support, solution pH, ionic strength, and presence
2.4 Catalyst Dispersion Considerations 41
Drying
The drying process can affect metal particle size, primarily when weak support–
precursor interaction exists. In this case, catalyst particles are formed by crystal-
lization of the precursor from the pore-filling liquid during solvent evaporation.
The ratio of nucleation rate to crystal growth rate determines the crystal size; if
it is large, the crystal size is small (Le Page, 1987). A high internal surface area
favors heterogeneous nucleation. The nucleation–crystallization process is influ-
enced by mass and heat transfer in the porous structure. When evaporation of the
42 Optimization of the Catalyst Distribution in a Single Pellet
Calcination
The calcination step, which is usually carried out in oxidizing atmosphere, can lead
to a number of transformations (Che and Bennett, 1989): precursor decomposition
and formation of an oxide species, bonding of the formed oxide with the support,
removal of some of the elements introduced during the preparation, decomposi-
tion of the precursor ionic complex, ligand exchange reactions between surface
anchoring groups and ligands bound to the metal ion, elimination of carbonaceous
impurities, and sintering of the precursor compound or the formed oxide species.
For platinum–silica catalysts, crystallite size was found to increase with calci-
nation temperature (Dorling et al., 1971; Brunelle et al., 1976). When catalysts
were obtained by chloroplatinic acid impregnation (no support–precursor inter-
action), the increase was more substantial than with catalysts prepared by amine
complex (interaction through ion exchange). Furthermore, dispersion loss started
at lower temperature for the former type of catalysts. For Pt/γ -Al2 O3 catalysts,
Bournonville et al. (1983) found that the metal dispersion exhibited a maximum as
a function of calcination temperature. It was proposed that increasing calcination
temperature facilitated the formation of an oxochlorinated platinum complex,
which starts to decompose above a certain temperature. The formation of this
complex was crucial for obtaining high dispersion. This complex was strongly
bound to the carrier and led to the formation of a well-dispersed metallic phase
upon reduction. Catalysts that were reduced without previous calcination had
2.4 Catalyst Dispersion Considerations 43
Reduction
Reduction in hydrogen is commonly employed for catalyst activation, during
which the metal precursor compound or its oxide is transformed into the metallic
state (Foger, 1984). The interaction of the oxide with the support, which depends
on the phenomena that have taken place during the preceding calcination step,
can also affect the reduction step (Delmon and Houalla, 1979). Various effects of
reduction temperature on particle size have been observed.
Increasing the temperature of reduction was found to result in an increase in
particle size for Pt/γ -Al2 O3 catalysts prepared by chloroplatinic acid impregnation
(Wilson and Hall, 1970). Only a small effect of reduction temperature on platinum
particle size was observed for catalysts prepared by a platinum amine complex ad-
sorption and reduced at 400–700◦ C (Brunelle et al., 1976). For Ni/Al2 O3 catalysts,
the nickel surface area showed a rather weak dependence on reduction temper-
ature between 300 and 500◦ C, with a maximum occurring at about 350–400◦ C.
Substantial loss of nickel surface area was obtained if the samples were calcined
before reduction (Bartholomew and Farrauto, 1976). Other treatments before re-
duction can also affect catalyst dispersion. Sarkany and Gonzalez (1983) showed
that initial pretreatment in helium results in Pt/γ -Al2 O3 catalyst with dispersion
considerably larger than that obtained after reduction. In this case, the dispersion
reached a value of 60% at loading of 2 wt% and remained constant up to 6 wt%
after helium pretreatment, whereas after reduction it decreased from 50% to 20%
as the loading increased from 2 to 6 wt%.
for various catalysts such as Pt/SiO2 (Dorling et al., 1971; Brunelle et al., 1976),
Pt/Al2 O3 (Basset et al., 1975), Rh/SiO2 (Arakawa et al., 1984; Underwood and
Bell, 1987), Rh/Al2 O3 (Fuentes and Figueras, 1980), Pd/SiO2 (Moss et al., 1979),
Pd/Al2 O3 (Scholten and Van Montfoort, 1962; Aben, 1968), Ag/SiO2 (Seyed-
monir et al., 1990), Ag/Al2 O3 (Gavriilidis et al., 1993b), Ni/Al2 O3 (Bartholomew
et al., 1980; Huang and Schwarz, 1987), Co/Al2 O3 (Reuel and Bartholomew, 1984),
Ir/Al2 O3 (Barbier and Marecot, 1981), Ru/Al2 O3 (King, 1978), and V2 O5 /ZrO2
(Chary et al., 1991).
A relationship between active element surface area per unit weight of catalyst
pellet, A, and weight fraction of active catalyst, q, which can be used to represent
many real situations is
pq
A= (2.122)
1 + bq
where p represents the specific surface area of active catalyst (square meters per
gram of active catalyst) and p/b is surface area of active catalyst per unit weight of
catalyst pellet at saturation, i.e. at high loadings. Note that from the active element
surface area A, the catalyst dispersion D, defined as the fraction of active element
atoms available on crystallite surface, is obtained as
D 1
= (2.123)
D0 1 + bq
impregnation (Gavriilidis et al., 1993b) are shown in Figure 2.21. The nonlinear
dependence of surface area on loading is due to the fact that α-Al2 O3 is an inert
support. In Figure 2.22, catalyst dispersion measured by hydrogen chemisorption
as a function of loading is shown for Rh supported on SiO2 . The catalysts were
prepared by incipient wetness impregnation of SiO2 , using aqueous solutions of
Rh(NO3 )3 (Underwood and Bell, 1987).
(2.122) the relation between reaction rate and catalyst loading is obtained:
pq
r = r (2.125)
1 + bq
where α is the maximum allowable value of the local active element concentration
on the support.
The reaction rate can be represented using equations (2.125) and (2.126) as
follows:
µ(s)
r = r0 . (2.129)
1 + Bµ(s)
The factor r0 represents the rate of reaction in a catalyst pellet with maximum
dispersion, i.e. where the active element surface area per unit catalyst weight is
equal to pq̄. The dimensionless quantity B = bq̄ is a measure of the nonlinearity
of the A-vs-q relation; B = 0 corresponds to a linear relation, while for increasing
B the nonlinearity becomes more severe. For a fixed catalyst preparation method
(i.e. fixed b value), B increases with the average active element loading of the
pellet, q̄.
As in section 2.3.2, we consider a general reacting system, where any number
of reactions occur, following arbitrary kinetics, in the presence of external mass
and heat transport resistances. The mass and energy balances in matrix form are
1 d dx µ(s)
L[x] = n sn = w(x) (2.130)
s ds ds 1 + Bµ(s)
48 Optimization of the Catalyst Distribution in a Single Pellet
with BCs
dx
s = 0: =0 (2.131a)
ds
dx
s = 1: = A (xf − x) (2.131b)
ds
where the vectors x, xf , and w and matrix A are defined in (2.101) and (2.102).
The effectiveness factor η M is defined as the ratio between the actual rate of con-
sumption of reactant M and the corresponding rate in the absence of all transport
resistances in a catalyst pellet with the same overall loading of active element
distributed uniformly throughout the support [i.e. µ(s) = 1]. Note that the pellet
used as reference in this definition is uniform while that under examination is
nonuniform, so they may have different values of the overall active surface area.
Thus
1
(n + 1)(1 + B) µ(s)
ηM = J w M (x)s n ds. (2.132)
DM j=1 ν Mj φ j 0
2 1 + Bµ(s)
dh
= µ(s)s n (2.137)
ds
2.5 Optimal Distribution of Catalyst Loading 49
where the last equation arises from the integral constraint (2.127). The corre-
sponding BCs follow from equations (2.127), (2.131), and (2.135):
s = 0: γ=0 (2.138a)
s = 1: γ = A(xf − x) (2.138b)
s = 0: h=0 (2.138c)
1
s = 1: h= . (2.138d)
n+1
When the effectiveness factor (2.132) is the objective functional, the Hamiltonian
is defined by (cf. Appendix A)
µ(s)
H= w M (x) + λT2 w(x) s n + λT1 γs −n + λ3 µ(s)s n (2.139)
1 + Bµ(s)
where λ1 , λ2 (column vectors, with I + 1 elements), and λ3 are the Lagrange
multipliers defined by the following differential equations:
dλ1 µ(s)
=− s n [w Mx + JT λ2 ] (2.140)
ds 1 + Bµ(s)
dλ2
= −λ1 s −n (2.141)
ds
dλ3
=0 (2.142)
ds
where w Mx is the column vector of derivatives of w M with respect to x, and J is
the Jacobian matrix of vector w with respect to the independent variable x:
∂w M /∂ x1
∂w M /∂ x2
∂w M ..
w Mx = = . (2.143)
∂x
∂w M /∂ x1
∂w M /∂ xI+1
∂w1 /∂ x1 ∂w1 /∂ x2 . . . ∂w1 /∂ xI ∂w1 /∂ xI+1
∂w2 /∂ x1 ∂w2 /∂ x2 . . . ∂w2 /∂ xI ∂w2 /∂ xI+1
.. .. .. ..
J= . . . . . (2.144)
∂w I /∂ x1 ∂w I /∂ x2 . . . ∂w I /∂ xI ∂w I /∂ xI+1
∂w I+1 /∂ x1 ∂w I+1 /∂ x2 . . . ∂w I+1 /∂ xI ∂w I+1 /∂ xI+1
The relevant BCs are
s = 0: λ1 = 0 (2.145a)
s = 1: λ1 = AT λ2 . (2.145b)
When the yield YNM is considered as the objective function, the same expres-
sions for the Hamiltonian and the Lagrange multipliers are obtained, but w M (x)
is now replaced by w N (x) (cf. Appendix A).
50 Optimization of the Catalyst Distribution in a Single Pellet
In the case where the objective functional is the selectivity SNM , it is convenient
to introduce two auxiliary variables xN and xM defined as follows:
dxN µ(s)
= w N (x)s n (2.146)
ds 1 + Bµ(s)
dxM µ(s)
= w M (x)s n (2.147)
ds 1 + Bµ(s)
with initial conditions
s = 0: xN = 0 (2.148a)
s = 0: xM = 0 (2.148b)
where the Lagrange multipliers are defined by the differential equations (2.141)
and (2.142) together with
dλ1 µ(s)
=− s n (JT λ2 + λ4 w Nx + λ5 w Mx ) (2.151)
ds 1 + Bµ(s)
dλ4
=0 (2.152)
ds
dλ5
=0 (2.153)
ds
along with BCs (2.145) and
DM
s = 1: λ4 = (2.154)
DN xM
DM xN
s = 1: λ5 = − 2
. (2.155)
DN xM
As explained in Appendix A, the solution of the optimization problem is given
by the active element concentration µ(s) which maximizes the Hamiltonian H
defined by equation (2.139) or (2.150), depending on the specific performance
index considered as the objective functional, while simultaneously allowing for
the constraint (2.128). The solution is described below for four cases (Baratti et al.,
2.5 Optimal Distribution of Catalyst Loading 51
1993, 1997). The first is the case of a single first-order isothermal reaction where
an analytical solution can be obtained, the second the case of linear dependence
between the active element surface area and its concentration, and the third the
case of a single first-order nonisothermal reaction where a numerical method of
general applicability is used for the solution. In the fourth case, the same method is
applied to investigate complex reacting systems which give rise to unusual optimal
loading distributions.
where interphase mass transfer resistance has been neglected. Following the gen-
eral procedure outlined above, the original problem (2.156) can be rewritten as
follows:
du
= γ s −n (2.158)
ds
dγ µ(s)
= φ2 s nu (2.159)
ds 1 + Bµ(s)
dh
= µ(s)s n (2.160)
ds
with the corresponding conditions
s = 0: γ =0 (2.161a)
s = 1: u=1 (2.161b)
s = 0: h=0 (2.161c)
1
s = 1: h= . (2.161d)
n+1
Using the effectiveness factor as the objective functional, the Hamiltonian is given
by (2.139):
µ(s)
H= (1 + λ2 )φ 2 us n + λ1 γ s −n + λ3 µ(s)s n (2.162)
1 + Bµ(s)
52 Optimization of the Catalyst Distribution in a Single Pellet
s = 0: λ1 = 0 (2.166a)
s = 1: λ2 = 0. (2.166b)
According to the maximum principle, in the case where the optimization function
is not bounded [i.e., the constraint (2.128) is not included, or α = ∞], the necessary
condition for the Hamiltonian H to be maximum leads to
∂H 1
= (1 + λ2 )φ 2 us n + λ3 s n = 0. (2.167)
∂µ [1 + Bµ(s)]2
By comparing equations (2.158) and (2.159) with equations (2.163) and (2.164),
together with the corresponding BCs (2.161a), (2.161b) and (2.166a), (2.166b), we
obtain
u = 1 + λ2 (2.168)
γ = −λ1 (2.169)
which may be verified by direct substitution. Using equation (2.168), the optimality
condition (2.167) reduces to
√
u −λ3
= (2.170)
1 + Bµ(s) φ
which, substituted in equations (2.158)–(2.160), leads to a system of three ODEs
with four BCs (2.161a–d). This set can be solved analytically to yield the three
functions u(s), γ (s), and h(s) in addition to λ3 , which according to equation (2.165)
is constant. The analytical expressions for the optimal active element distribution
µ(s) , the reactant concentration profile u(s), and the catalyst pellet effectiveness
factor η are reported in Table 2.4 for three pellet geometries: slab (n = 0), cylinder
(n = 1), and sphere (n = 2).
In Figure 2.23.a, the optimal distribution of the active element µ(s) is shown √
for a sphere and for various values of the dimensionless parameter ψ = φ/ B,
which can be regarded as a modified Thiele modulus that takes into account the
nonlinearity in the dependence of the active surface area upon the active element
concentration. Note that this is the only variable needed to fully characterize the
optimal distribution. For values of ψ approaching 0, the optimal distribution ap-
proaches the uniform distribution. This is physically sound, considering that when
the concentration of the reactant inside the pellet is uniform (small φ values) or
2.5 Optimal Distribution of Catalyst Loading 53
Table 2.4. Optimal active element distribution, with the corresponding reactant
concentration profile and the effectiveness factor,
√for three pellet geometries and for
a single first-order isothermal reaction (ψ = φ/ B).
when the nonlinearity in the A-vs-q dependence is strong (large B values), the cat-
alyst is more effectively utilized when it is distributed as uniformly as possible. On
the other hand, when ψ → ∞, the optimal distribution becomes a progressively
steeper monotonically increasing function of s, exhibiting its maximum value at
the pellet external surface. This occurs when either B → 0 or φ → ∞. For B → 0,
since there is no problem of loss of catalyst surface area with increasing loading,
it is better to locate all the catalyst where the reactant concentration is maximum,
i.e. at the external surface of the pellet. For increasing values of φ, no matter how
large B is, the reaction takes place in a narrower region near the pellets external
surface, which is the only place where the active element can be profitably located.
Figure 2.23. (a) Optimal active element distribution for various values of
the modified Thiele modulus ψ, and (b) corresponding values of the effec-
tiveness factor. Reaction system: a single isothermal first-order reaction
in a sphere (n = 2). (From Baratti et al., 1993.)
54 Optimization of the Catalyst Distribution in a Single Pellet
This can be seen explicitly from the expressions for the optimal µ(s) reported in
Table 2.4, where for any of the examined geometries, in the limit as ψ → ∞,
ψ
µ(s) → exp[−ψ(1 − s)]. (2.171)
(n + 1)s n/2
so that the η-vs-φ dependence is similar to that for uniform pellets (cf. Aris, 1975).
2.5.3 Linear Dependence between the Active Element Surface Area and Its Loading
When the dependence between the active element surface area and its loading
is linear, i.e. B = 0, the expressions of both Hamiltonians (2.139) and (2.150) are
linear with respect to µ, and can be reduced to the following form:
H = µ + ω (2.173)
The function is called switching function, and its expression depends on the reac-
tion system under consideration as well as the performance index to be optimized.
Both and ω are independent of µ. In particular, considering the performance
indices (2.132)–(2.134), the following expressions for the switching function are
obtained:
= w M (x) + λT2 w(x) + λ3 s n for η M (2.174)
= w N (x) + λT2 w(x) + λ3 s n for YNM (2.175)
T n
= λ2 w(x) + λ3 + λ4 w N (x) + λ5 w M (x) s for SNM (2.176)
case, the Hamiltonian is maximized by setting the control variable at its maximum
value if the switching function is positive and at its minimum value if the switching
function is negative. Hence, in general, the optimal distribution µ(s) is given by a
multiple step function, where the positions and number of steps are controlled by
the number of sign changes of the switching function . Accordingly, the optimal
active element concentration profile µ is given by
α when > 0
µ= (2.177)
0 when < 0.
For example, in the simple case of a single first-order isothermal reaction with
the effectiveness factor as the objective functional treated in section 2.5.2, the
switching function can be readily obtained from equations (2.162) and (2.168):
= (u2 φ 2 + λ3 )s n . (2.178)
Figure 2.26. Ratio of effectiveness factors for optimal and uniform distributions. Reaction
system: a single, nonisothermal, first-order reaction, with γ = 18, Bim = ∞, Bih = ∞, and
n = 2. (From Baratti et al., 1993.)
(a) (b)
can be a multiple-step function. This was demonstrated recently for the case of
two parallel, irreversible reactions (Baratti et al., 1997)
1
A −→ B
2
(2.180)
A −→ C
um j exp[γ j (1 − 1/θ )]
f j (u, θ) = , j = 1, 2. (2.181)
{1 + σ j u exp[ε j (1 − 1/θ )]}2
In the case of negligible external mass and heat transfer resistances and equal
heats of reaction, Prater’s relationship between the dimensionless reactant con-
centration u and dimensionless temperature θ leads to [see also equation (2.43)]
This relation allows one to represent the overall dimensionless rate of consumption
of reactant A as a function of only one variable:
k 2,f
f (u) = f1 [u, 1 + β(1 − u)] + f2 [u, 1 + β(1 − u)]. (2.183)
k 1,f
The behavior of the global reaction rate f (u) is shown in Figure 2.28 for three
values of k2,f /k1,f , which is the ratio of the reaction constants at bulk fluid temper-
ature. These values were chosen for illustrative purposes, so as to clearly separate
the two maxima of function f. It is assumed that the dependence of catalyst activ-
ity on loading is linear (i.e. B = 0). We seek to maximize the overall effectiveness
factor
n+1 1
η= µ(s) f (u)s n ds (2.184)
f (1) 0
while satisfying the constraints (2.127) and (2.128). Equations (2.135)–(2.137) now
become
du
= γ s −n (2.185)
ds
dγ
= φ 2 µ(s) f (u)s n (2.186)
ds
dh
= µ(s)s n . (2.187)
ds
The corresponding BCs are
s = 0: γ =0 (2.188a)
s = 1: u=1 (2.188b)
s = 0: h=0 (2.188c)
1
s = 1: h= . (2.188d)
n+1
Using the overall effectiveness factor as the objective functional, from equation
(2.139) the Hamiltonian is given by
= [(1 + λ2 )φ 2 f + λ3 ] s n . (2.190)
s = 0: λ1 = 0 (2.194a)
s = 1: λ2 = 0. (2.194b)
Since the Hamiltonian is linear with respect to µ(s), the optimal catalyst loading
profile is the following step function (see also section 2.5.3):
α when > 0
µ= (2.195)
0 when < 0.
The number and location of steps depend on the specific form of the switching
function, whose evaluation requires complete solution of the problem. This has
62 Optimization of the Catalyst Distribution in a Single Pellet
been carried out numerically, using the method described in section 2.5.4, and the
results are shown in Figure 2.29 for a few selected values of the upper bound α on
the dimensionless catalyst loading.
For low values of α, the optimal distribution is given by a single-step function.
As α increases, it becomes possible to concentrate more catalyst in a given support
area. In particular, for the cases shown in Figure 2.29.c and d, the catalyst has been
placed in two locations, which (as can be seen from the corresponding average
values of u) are near to those where f (u) exhibits its maxima. In these cases, the
optimal distribution is given by a two-step function. When α is further increased
(α > 7.4), the optimal distribution again takes the form of a single-step function,
since α is now large enough so that all the catalyst can be located in the vicinity
of the point where its utilization is maximized, i.e. where f (u) exhibits the larger
of its maxima. In the present case (see Figure 2.28), this corresponds to the lower
value of u, and hence to the inner of the two steps noted above. In the limit as
α → ∞, the optimal distribution approaches a Dirac-delta function.
2.6 Experimental Studies 63
which were obtained under excess oxygen are, consistent with the detrimental ef-
fect of temperature gradients on selectivity (Morbidelli et al., 1984a; Pavlou and
Vayenas, 1990a), owing to the higher activation energy of ethylene combustion
than of ethylene epoxidation. Under ethylene-rich conditions, in addition to tem-
perature gradients, oxygen concentration gradients affect the selectivity negatively
(Yeung et al., 1998). Reactant concentration gradients are also detrimental for
Eley–Rideal rate equation, which is of positive order with respect to both reac-
tants. Five different catalyst distributions ranging from preferential shell loading
to preferential core loading were examined under isothermal conditions. The ef-
fectiveness factor was found to be higher for pellets with the catalyst concentrated
towards the external surface, which is consistent with the Eley–Rideal kinetics.
At small Thiele modulus (i.e. kinetic control), all pellets showed the same perfor-
mance as expected, but the shell-type pellets showed notable improvements for
larger Thiele modulus values.
In complex hydrogenation reactions, selectivity is the performance index of
interest. In these cases, partially hydrogenated products are typically desired.
Komiyama et al. (1997) studied liquid-phase batch hydrogenation of 1,3-butadiene
over Pt/γ -Al2 O3 catalysts, which is a triangular reaction network
Butadiene → Butene
(2.196)
Butane
I
n single-pellet studies, it is assumed that no concentration or temperature gradi-
ents are present in the fluid phase surrounding the pellet. A reactor with external
or internal recycle is one of the experimental realizations of the single-pellet
concept. However, in a fixed-bed reactor, the fluid-phase composition and temper-
ature vary with position. For this reason, the optimization problem becomes more
complex. Thus, it is not surprising that relatively few reactor studies have appeared
in the literature as compared with those for single pellets. In this chapter, we first
discuss theoretical studies of single and multiple reactions under isothermal and
nonisothermal conditions, and then present experimental work which supports
the theoretical developments.
69
70 Optimization of the Catalyst Distribution in a Reactor
z = 0: C f = Cf0 (3.2)
where r̄ (Cf ) represents the volume-averaged reaction rate in the catalyst phase
and is expressed by
1
r̄ (Cf ) = r (C) a(x) dVp . (3.3)
Vp Vp
y = 0: g=1 (3.5)
r (C) (1 + σ )2 u (3.6)
u = C/Cf0 , σ = KCf0 , f (u) = = .
0
r (Cf ) (1 + σ u)2
3.1 A Single Reaction 71
The fluid-phase balance has to be solved in conjunction with the solid-phase bal-
ance [equations (2.2) and (2.5)], which using the above dimensionless variables
reduces to
1 d n du
s = φ 2 a(s) f (u) (3.7)
s n ds ds
with BCs
du
s = 0: =0 (3.8a)
ds
s = 1: u=g (3.8b)
where the second BC allows for the fact that the fluid-phase composition changes
along the reactor length.
The objective of reactor optimization is to maximize the conversion at the
reactor outlet,
X = 1 − g(1). (3.9)
where
1
e(y) = (n + 1) f (u)a(s, y)s n ds (3.11)
0
δ[s − s̄(y)]
a(s, y) = (3.12)
(n + 1)s̄ n (y)
72 Optimization of the Catalyst Distribution in a Reactor
σ < 1 or
σ > 1 and g < 1/σ σ > 1 and g > 1/σ
s̄opt (y) 1 1 − a n=0
exp(−) n=1
1
n=2
1+
(1 + σ )2
eopt (y) f (g) f (um ) =
4σ
a For > 1, s̄opt = 0 and eopt (y) = f (ū), where ū is the solution of 1 − ū −
φ 2 f (ū) = 0.
where
4(n + 1)(σ g − 1)
= (3.14)
φ02
so that, since g = g(y), s̄opt is a function of reactor position y. The resulting value of
the local consumption rate, e(y), is obtained by substituting the optimal Dirac dis-
tribution in equation (3.11). Note that these solutions do not hold for the cases σ <
1, or σ > 1 and g < um = 1/σ . In these cases the reaction rate f (u) is monoton-
ically increasing in the interval u ∈ [0, g] and attains its maximum value at u = g.
Therefore, the optimal catalyst location is at the pellet external surface, i.e. s̄opt = 1.
Note also that for the infinite slab (n = 0), the value of s̄opt can become negative.
In this case, as discussed in section 2.1.1, the optimum catalyst location is s̄opt = 0.
The expressions for the optimal location along the reactor and the correspond-
ing outlet reactant conversion are summarized in Table 3.1. They indicate that,
starting from the reactor inlet, until the fluid reactant concentration reaches 1/σ ,
the optimal pellets have subsurface Dirac-delta distribution. The catalyst location
moves progressively towards the pellet external surface, while each pellet exhibits
constant local reaction rate, eopt .
When the fluid reactant concentration becomes lower than 1/σ , then beyond
this reactor position, the pellets have surface Dirac distribution and the local
reaction rate changes with fluid reactant concentration. The fluid concentration
g and the optimal location s̄opt are illustrated in Figure 3.2 for an optimal reactor
3.1 A Single Reaction 73
Figure 3.2. Rigorous optimal solution: profiles of the optimal active catalyst
location s̄ opt and the bulk fluid concentration g along the reactor axis y for
a bimolecular Langmuir–Hinshelwood reaction. (From Morbidelli et al.,
1986b.)
where g(y) > 1/σ . Integrating equation (3.4) yields the fluid concentration profile
Note that in this case, only the optimal catalyst location s̄opt depends on the Thiele
modulus, while the g profile remains unchanged, since diffusional resistances affect
only the location inside the pellet where um is realized.
The obtained optimal solution requires the catalyst location to be different in
each pellet, depending on its location along the reactor. This is difficult to realize
in practice, but provides the theoretical limit of reactor performance, and hence
may be referred to as the absolute optimum solution. The more practical situa-
tion where the catalyst location is the same for all pellets has also been examined
(Lee et al., 1987a). Its performance was only slightly below the absolute optimum
solution for low conversion, but became progressively worse for higher conver-
sions. As an alternative configuration, the fixed-bed is divided into multiple zones,
where in each zone the catalyst location is the same but different from the others.
The optimization variables are now the catalyst location within pellets in each
zone, as well as the length of the corresponding zone. As shown in Figure 3.3,
for any given Damköhler number Da, and for increasing number of zones, d, the
optimized reactor configuration evolves towards the absolute optimum, achieved
for an infinite number of zones, d = ∞. By the use of only two or three zones, the
theoretical optimal reactor performance is reasonably closely approached. The
optimal catalyst locations and zone lengths for reactors having one, two, three, or
an infinite number of zones are illustrated in Figure 3.4.
In reactor optimization studies, care should be exercised for the possible oc-
currence of multiple steady states. For the reaction system considered above,
74 Optimization of the Catalyst Distribution in a Reactor
reactor steady-state multiplicity was observed (three steady states, of which two
were stable) and attributed to the multiplicity behavior of the first layer of pellets
packed along the axial direction (Morbidelli et al., 1986a). Criteria were developed
to predict the existence of multiple steady states and ignition conditions. In the
case of a single-zone reactor with axial dispersion, the number of possible steady
states increases to nine, among which only three are stable (Lee et al., 1987b).
Two sources of multiplicity are present in this case: the heterogeneous nature of
the reactor and the presence of axial dispersion in the fluid phase.
Homogeneous–heterogeneous reactions
In the cases discussed up to now, only catalytic reactions have been considered.
However, there are certain situations where both heterogeneous and homoge-
neous reactions take place, with some species participating in both types of re-
actions. One such case is catalytic combustion, where heterogeneous reactions
can enhance homogeneous ones because the catalyst promotes the production of
free radicals, which can increase the rate of homogeneous reactions (Pfefferle and
Pfefferle, 1987). The simple case of two first-order reactions, one heterogeneous
and the other homogeneous, was studied by Melis et al. (1997). The analysis was
carried out in the absence or presence of interactions between the two reaction
Figure 3.4. Optimal catalyst location s̄ opt along the reactor axis y for various num-
bers of zones in the reactor, d. Bimolecular Langmuir–Hinshelwood reaction, σ =
5, Da = 0.3, φ 2 = 0.7, n = 1. (From Lee et al., 1987a.)
3.1 A Single Reaction 75
z = 0: Cf = Cf0 , Tf = Tf 0 . (3.18)
Figure 3.5. Reactor exit conversion X as a function of dimensionless length of the first
(catalytic) zone when the homogeneous reaction rate in the second (noncatalytic) zone is
enhanced through the catalytic reactions by the enhancement parameter shown. Da = 1.
Dimensionless mass transfer coefficient, dimensionless heterogeneous reaction rate con-
stant: (a) 1, 1; (b) 1, 5; (c) 5, 1; (d) 5, 5. (From Melis et al., 1997.)
y = 0: g = 1, τ =1 (3.21)
The above fluid-phase balances need to be solved in conjunction with the solid-
phase balances (2.43)–(2.45), which using the above dimensionless variables take
3.2 Multiple Reactions 77
the form
1 d du
sn = φ 2 a(s, y) f (u, θ) (3.23)
s n ds ds
1 d n dθ
s = −βφ 2 a(s, y) f (u, θ ) (3.24)
s n ds ds
du dθ
s = 0: = 0, =0 (3.25a)
ds ds
s = 1: u = g, θ = τ. (3.25b)
A1 → A2
(3.26)
A1 → A 3
where A2 is the desired product. The kinetic expressions considered for the two
reactions are
k1 C1m1
r1 = (3.27a)
(1 + K1 C1 )b1
k2 C1m2
r2 = . (3.27b)
(1 + K1 C1 )b2
78 Optimization of the Catalyst Distribution in a Reactor
dgi 2
= νij Da j f j (ū). (3.29)
dy j=1
Since the reaction rates (3.27) depend only on the concentration of a single com-
ponent, the vector u is in reality simply u1 .
The reactor selectivity and yield are defined by
g2 (1)
S= (3.30)
1 − g1 (1)
Y = g2 (1) = X S (3.31)
where X is the reactor conversion. According to the local OCD approach, the
optimal value of ū1 is determined as the one which locally maximizes the desired
reactor performance. In the case where such a value of ū1 is larger than the cor-
responding concentration value in the bulk, g1 , the optimal performance is not
obtainable and we have the suboptimal situation where ū1 = g1 and s̄ = 1. It is
in this event that the global OCD approach can provide better performance. In
this case the ū1 value is obtained numerically by a one-parameter optimization
procedure where now the objective function is not the local but rather the global
reactor performance. In Figure 3.6, with reference to the reacting system above, it
is shown that the global OCD approach can result in higher reactor yield than the
local approach. A difference in the performance index obtained by the two meth-
ods exists only when the reactor has two sections: one with s̄opt < 1 (0 < y < ŷ)
and the second with s̄opt = 1( ŷ < y < 1). The difference increases when the length
of the second section (1 − ŷ) increases. Similarly, the global OCD approach can
also give higher reactor selectivity than the local OCD approach (Wu, 1994).
An alternative design approach, which utilizes a one-zone reactor of Dirac cat-
alysts but still results in the same optimum performance obtained by the global
OCD approach, has been proposed (Sheintuch et al., 1986; Wu, 1995). This was
accomplished by optimizing the feed distribution along the reactor. In particu-
lar, Wu (1995) considered the case of two isothermal parallel reactions following
Langmuir–Hinshelwood kinetics (3.27), with yield as the performance index. Uti-
lizing the maximum principle, the optimal feed distribution was found analytically
to be a uniform distribution of the feed stream along a reactor portion starting
from its entrance, so the optimization problem was reduced to determining the
length of this portion and the optimal catalyst location. For low Damköhler val-
ues, the feed distribution portion was the whole reactor length and the associated
optimal catalyst location was a subsurface position. For large Damköhler values,
the feed distribution portion was shorter than the reactor length and the optimal
catalyst location was the pellet external surface.
3.2 Multiple Reactions 79
A comparison of the optimal feed distribution (OFD) and the global OCD
approaches is shown in Figure 3.7. Both approaches lead to the same outlet yield,
and the length of the feed distribution for the OFD approach is equal to the
length of the reactor portion with subsurface optimal catalyst location for the
OCD approach. These similarities are not coincidental and are directly related to
the fact that the ū1 profile along the reactor is the same. For the OCD approach
this profile is achieved by adjusting the catalyst location in pellets at each point
along the reactor; for the OFD approach it is achieved by properly distributing
the feed. Therefore the two approaches give the same reactor performance and
differ only in the way they achieve it.
k1 C1 C32
r1 = √ (3.33a)
(0.0106 + 2144C1 + 805C3 )(1 + 1271 C3 )
k2 C1 C32
r2 = √ (3.33b)
(0.008 + 4166C1 + 1578C3 )(1 + 718 C3 )
where C1 and C3 are the concentrations (mol/cm3 ) of ethylene and oxygen, re-
spectively. This is a parallel reaction scheme similar to (3.26). Yield and selectivity
80 Optimization of the Catalyst Distribution in a Reactor
used as the performance indexes are given by equations (3.30) and (3.31), where
g1 and g2 are the dimensionless bulk fluid concentrations of ethylene and ethy-
lene oxide, respectively. The fluid-phase mass and energy balances for a one-zone
reactor are
2 1
dgi
= νij Da j (n + 1) f j (u, θ ) a(s)s n ds (3.34)
dy j=1 0
2 1
dτ
= β̄ j Da j (n + 1) f j (u, θ ) a(s)s n ds − St(τ − τc ) (3.35)
dy j=1 0
dgi 2
= νij Da j f j (ū, θ̄ ) (3.36)
dy j=1
dτ 2
= β̄ j Da j f j (ū, θ̄) − St (τ − τc ) (3.37)
dy j=1
3.2 Multiple Reactions 81
where St is the Stanton number. The solid-phase mass and energy balances (2.94)–
(2.96) were reduced to algebraic equations using standard techniques. Optimiza-
tion was carried out numerically, through one- or multiparameter estimation meth-
ods for the one- and two-zone reactors, respectively.
For small values of the Thiele modulus, the optimal catalyst location for yield
maximization was mostly in the pellet interior. In this case, no discernible dif-
ference in the performance of the optimal, surface, and uniform distributions
was observed, because kinetic control prevailed. Increasing catalyst activity led to
ignition behavior for both surface and uniform catalysts. For the optimal Dirac dis-
tribution, though, intraparticle gradients present for subsurface catalysts helped
to avoid reactor runaway. Thus, when using Dirac catalysts it is possible to operate
the reactor with higher loading (activity) catalysts, while avoiding reactor runaway.
This has a beneficial effect on reactor performance, since while ignition results in
low yields for the uniform and surface distributions, the yield obtained with the
optimal Dirac catalyst continues to increase with catalyst activity. This behavior is
clearly demonstrated in Figure 3.8, where dimensionless bulk-fluid concentrations
and temperature rise (above the inlet reactor temperature) along the reactor are
shown for the three distributions, using parameter values representative of typical
reaction conditions. The surface pellet ignites earlier than the uniform, because
there are no intraphase concentration gradients to slow down reaction rates. The
runaway situation is not observed for the optimal Dirac distribution. In this case,
there is a hot spot close to the reactor entrance, but eventually the temperature
stabilizes.
For selectivity maximization only two positions were found to be optimal, de-
pending primarily on the value of Thiele modulus: surface and center. This behav-
ior is directly related to the complexity of the intrinsic reaction kinetics. However,
subsurface locations were associated with negligible ethylene conversion, so that
the surface distribution is the only one of interest from the practical point of
view.
In Figure 3.9, a comparison between one- and two-zone reactors is shown for
yield maximization. The optimal catalyst location in the second zone, s̄2,opt , is larger
than in the first, s̄1,opt , in order to compensate for the decreased bulk reactant
concentration. The optimal location s̄opt for the one-zone reactor falls between
s̄1,opt and s̄2,opt , since it is in a sense their average. As the catalyst activity increases,
the optimal location for the second zone reaches the pellet surface. It does not
move to subsurface locations as s̄opt and s̄1,opt do, because in the second zone
reactant concentrations are low, thus preventing runaway. The improvement by
using a two-zone reactor is insignificant at low values of Thiele modulus, but at
larger values yield improvements of about 15% over the single-zone reactor can
be realized.
The fact that optimization of catalyst distribution leads to avoidance of reactor
runaway in ethylene epoxidation is attractive. Since the desired reaction has lower
activation energy than the undesired reaction, excessive temperatures result in low
selectivity and yield. Hence, in order to maximize yield and selectivity, reactor run-
away must be avoided. It is worth noting that axial temperature gradients can also
be suppressed by using uniform pellets and diluting them appropriately with inert
pellets or by using uniform pellets but with catalyst loading depending on their
position along the reactor. In this way, a nonuniform activity profile is established
along the reactor length. Both approaches have been shown to prevent reactor
runaway and can also lead to improved selectivity (Caldwell and Calderbank,
1969; Narsimhan, 1976; Sadhukhan and Petersen, 1976; Pirkle and Wachs, 1987).
of the pellet radius) and located approximately at s̄ = 0.27, 0.50, 0.65, 1. Two steady
states were observed in all single-zone reactors. As shown in Figure 3.10, among the
four types of catalysts studied, the reactor packed with the catalyst where the active
layer was positioned at s̄ = 0.65 had the widest range of high-conversion regime.
This range was further increased by using a dual-zone reactor, with s̄ = 0.27 and
s̄ = 1 in the first and second zones, respectively. This improvement was expected
according to the discussion in section 3.1.1. A direct experimental test of the
effect of nonuniform catalyst distribution was demonstrated in a dual-zone reactor.
Using the same catalyst pellets, the reactor performance (including the number of
steady states) was dramatically altered when the packing order was reversed, i.e.,
first s̄ = 0.27 and second s̄ = 1 in the two zones, or vice versa, as shown in Figure
3.11. All of the above observations were in excellent qualitative agreement with
theoretical predictions.
3.3 Experimental Studies 85
A
n important area where the techniques of catalyst design described above
can be used to improve process performance involves reacting systems whose
activity changes with time. These include catalysts undergoing poisoning as
well as processes where small amounts of undesired components are removed
from a stream (e.g. hydrodemetallation). In such catalytic systems a new variable
is introduced: the catalyst lifetime. It should be noted that only the initial catalyst
distribution can be controlled, while the performance index is usually integrated
over the catalyst lifetime. Next, we discuss catalyst design in the presence of two
deactivation mechanisms: nonselective and selective poisoning.
86
4.1 Nonselective Poisoning 87
du
s = 0: =0 (4.3a)
ds
s = 1: u=1 (4.3b)
sp = 1 − π (4.4)
At low Thiele modulus (φ → 0), it was found that the optimal distribution is a
Dirac-delta located at the pellet center. This is because under kinetic control,
any catalyst distribution provides the same performance if no deactivation occurs,
while the time required for the poison to reach the catalyst is maximum if the
catalyst is located at the pellet center.
For large Thiele-modulus, the optimal initial activity distribution tends to be
bimodal, concentrating one portion of the active element near the pellet’s external
surface and another portion close to its center (Pavlou and Vayenas, 1990b), as
illustrated in Figure 4.2.a for π = 1. The specific shape depends on the pellet
geometry, values of the physicochemical parameters involved, and kinetics of the
deactivating process through the value of the parameter π . The catalyst loading
required near the center increases in going from the slab to the sphere geometry,
Figure 4.2. Optimal catalyst distribution for nonselective poisoning. First-order main re-
action with (a) constant velocity of the deactivating front (π = 1), (b) diffusion-controlled
deactivation (π = 0.5). Here φ[(n + 1)/gv ]1/2 = 20. (From Pavlou and Vayenas, 1990b.)
4.2 Selective Poisoning 89
with a concomitant decrease near the external surface. When π = 0.5, the catalyst
loading is substantially smaller near the surface and in fact vanishes at the surface,
as shown in Figure 4.2.b. This behavior results from the interaction of two opposing
factors. One is the effectiveness factor, which dictates that the catalyst should be
placed close to the surface. The other is the catalyst lifetime, which dictates that
it should be placed near the center. As seen in Figure 4.2.a, these factors result in
a bimodal optimal distribution for π = 1. However, when π = 0.5, the poisoning
front moves faster at the beginning. Consequently, the optimal catalyst loading
near the pellet surface decreases substantially. Finally, it should be noted that for
the case examined, contrary to nondeactivating systems, depositing the catalyst
nonuniformly does not improve the pellet performance significantly.
For an isothermal first-order reaction, the case of uniform poisoning, in both
single pellets and fixed-bed reactors, was considered by DeLancey (1973). The
objective function included economic considerations related to conversion of re-
actants and catalyst replacement. An approximate solution was derived using
Pontryagin’s principle. For a single pellet the optimal catalyst distribution was
eggshell, while for a fixed-bed reactor it was uniform.
= 0: a = a(s, 0) (4.11)
An economic criterion based on profit per unit time was used as the performance
index, which takes into account price of product and cost of catalyst replacement:
α1 0 oper η() d − α3
G= (4.12)
oper
where oper is the dimensionless operating time and α1 , α3 are weighting coeffi-
cients proportional to product value and catalyst cost, respectively. The optimiza-
tion variables were the initial catalyst distribution a(s, 0) and the operating time.
By developing an optimality condition and following similar arguments to those
in section 2.3.2, it was shown that the optimal distribution is a Dirac-delta func-
tion if no upper bound is imposed on the local catalyst loading. In the case where
local loading is bounded, the optimal distribution is expected to take the form of a
step function, in analogy with nondeactivating systems described in section 2.5.3.
This result is valid for multiple reactions, following arbitrary kinetics, in the pres-
ence of thermal effects as well as inter- and intraparticle mass and heat transport
resistances.
The case of an isothermal first-order reaction where the poisoning kinetics is
also first-order was investigated as an example (Brunovska et al., 1990). It was
found that for low values of the ratio of the two Thiele moduli characteristic of
the poisoning and main reactions, φ p /φ, the optimal location is at the pellet exter-
nal surface, because the intraparticle resistance is larger for the reactant than for
the poison. For increasing values of this parameter, the optimal location moves
towards the pellet interior, to an extent which depends on the specific operating
conditions. Further, the optimal operating time increases monotonically with φp /φ.
The case of selective poisoning in a fixed-bed reactor was studied by Markos
and Brunovska (1989) for a nonisothermal plug-flow reactor. Optimization was
carried out only for Dirac-type distributions constituting a single-zone reactor, and
in all cases significant improvements were reported when using the appropriate
Dirac-delta distribution. First-order deactivation kinetics (on poison concentra-
tion) and first-order kinetics of the main reaction (on reactant concentration) were
considered. The objective function took account of the catalyst and raw-materials
cost as well as the value of the products. It was defined similarly to (4.12), but in
this case the integral of outlet reactor conversion was used
(α1 − α2 ) 0 oper X d − α3
K= (4.13)
oper
where α2 is a weighting coefficient proportional to the cost of raw materials.
Two competing factors determine the optimal catalyst location. Surface loca-
tions favor high initial reactor conversion, while internal locations are beneficial
for longer catalyst lifetime. This can be seen in Figure 4.3, where the objective
function is shown as a function of the operating time oper for various active
4.3 Experimental Studies 91
4.3.1 Methanation
The methanation reaction
over Ni/Al2 O3 was studied in step-type catalyst pellets by Wu et al. (1990b). This
reaction is accompanied by selective poisoning due to carbon deposition (coking)
when the H2 /CO ratio is low. The pellets were prepared by pressing an active pow-
der (Ni/Al2 O3 ) layer sandwiched between two inert powder (Al2 O3 ) layers. It was
found that significant coking occurred only when the active layer was located close
to the external surface of the pellet. By placing the catalyst layer deeper inside,
diffusional resistances introduced by the outer Al2 O3 layer increased the H2 /CO
ratio at the catalyst location, since the diffusivity of hydrogen is larger than that of
carbon monoxide in the Knudsen diffusion regime. High H2 /CO4 ratios suppress
coking, but if the catalyst is placed too deep inside the pellet, the effectiveness fac-
tor decreases significantly. Accordingly, for the reaction conditions investigated,
a pellet with the catalyst layer centered at s̄ ∼ 0.6 had the highest effectiveness
factor while simultaneously avoiding coking, as shown in Figure 4.4. Agreement
of experiments with theoretical diffusion–reaction calculations was good, as also
shown in the figure.
4.3.2 Hydrogenation
Ethylene hydrogenation in the presence of low concentrations of the poison thio-
phene on Pt/Al2 O3 catalysts was studied by Brunovska and coworkers (Pranda
and Brunovska, 1993; Remiarova et al., 1993; Brunovska et al., 1994). Pellets were
4.3 Experimental Studies 93
also showed smaller coke formation and slower deactivation. This was attributed
to the higher hydrogen partial pressure in the reaction zone, which inhibits coke
buildup and further polymerization. In addition, coke precursors could diffuse out
from the catalyst pores more easily, hence decreasing coke formation. However,
palladium migration from the outside towards the core of the pellet occurred for
eggshell catalysts (Lin and Chou, 1995), during hydrogenation as well as subse-
quent catalyst regeneration. This migration was accompanied by sintering, which
resulted in loss of catalytic surface area.
4.3.3 NO Reduction
Stenger et al. (1988) and Hepburn et al., (1991b) studied the design of Rh/Al2 O3
honeycomb catalysts for NO reduction with H2 , in the presence of SO2 , for use in
stationary pollution sources such as power plants and industrial boilers. Eggshell
and egg-white catalysts were prepared by coimpregnation of the support with
RhCl3 and HF. Electron probe microanalysis revealed that sulfur was distributed
throughout the support, indicating that SO2 was not limited by diffusion. Thus,
subsurface location of Rh does not delay deactivation. It is worth noting that
sulfur deposition on alumina was small, while higher sulfur concentrations were
observed where Rh was present, indicating selective poisoning. The egg-white
catalyst exhibited greater tolerance to deactivation than the eggshell catalyst,
especially at longer periods. This behavior was attributed to the presence of the
outer alumina layer, which scavenges the SO2 before it reaches the noble metal.
5
Membrane Reactors
M
embrane reactors offer the advantage over conventional fixed-bed reactors
of combining chemical reaction and separation in a single unit. They can
substantially improve the performance of reactions by selective removal of
one of the reaction products or by controlled addition of a reactant. In the former
case, conversion enhancement beyond the thermodynamic limit can be achieved
for equilibrium-limited reactions, while in the latter, product selectivity can be
improved by influencing the concentration and residence time of components
giving rise to undesired reactions. Several articles which review experimental and
theoretical studies of catalytic membrane reactors are available in the literature
(Hsieh, 1991; Tsotsis et al., 1993; Saracco and Speccia, 1994; Zaman and Chakma,
1994).
95
96 Membrane Reactors
mass balances for the three reactors were solved analytically for uniform and Dirac
catalyst distributions (see Szegner, 1997 for details). For the membrane reactors,
the total conversion XT is computed by combining the feed and permeate sides.
It is thus defined as the ratio between the total molar flow of reactant converted
to product and the initial molar flow of reactant on the feed and permeate sides:
ζF,A
0
+ ζP,A
0
"p0 − ζF,A "F − ζP,A "p
XT = (5.1)
ζF,A
0
+ ζP,A
0
"P0
where ζF,A , ζP,A , "F , and "P are the reactant mole fractions and dimensionless
molar flow rates on the feed and permeate sides, respectively, with
QF QP
"F = , "P = . (5.2)
QF0 QF0
The dimensionless residence time is defined as the ratio between residence time
and the characteristic time for diffusion:
Vr PF /Rg T QF0 De,A Vr PF
= = . (5.3)
δM /De,A
2
Rg T QF0 δM
2
5.1 Membrane Reactors with Nonuniform Catalyst Distribution 97
selective product separation does not play a role in enhancing the conversion. On
comparing results of the two membranes (i.e. Figure 5.2.a and b) it is seen that the
improvement in reactor performance on locating the Dirac catalyst at s̄ or ξ̄ = 1, as
compared to the uniform catalyst distribution, is greater for the CMR. This occurs
because in the case of the uniform CMR, as the reactant diffuses through the
membrane, the back reaction of the product competes with its selective removal.
In the case of the IMRCF, the difference in the performance between the uniform
and the surface distribution is not as large, because the separation occurs outside
the reaction zone.
A comparison of the various reactor configurations is shown in Figure 5.3.
The function of the membrane in this system is evident on considering that at
large residence time the fixed-bed performance is limited to XT = 0.5 by chemical
equilibrium. That limit is exceeded by all membrane reactors. The Dirac-delta dis-
tribution located at the feed side or at the pellet surface (i.e. ξ̄ = 1 for the CMR
and s̄ = 1 for the IMRCF and FBR) shows superior performance to a uniform
activity profile. This distribution maximizes the reactant concentration at the cat-
alyst site, since it is not limited by diffusion through the membrane or the pellet.
The performance of the CMR and IMRCF is identical when the Dirac location
is at the feed side of the membrane (for the CMR) or at the external surface
of the pellet (for the IMRCF), due to the absence of mass transfer resistances.
These two membrane configurations exceed the performance of the FBR con-
figurations over the full range of residence time. However, for small the FBR
with s̄ = 1 (i.e. Dirac-delta at the external surface of the pellet) performs better
than the IMRCF with a(s) = 1 (i.e. uniform catalyst distribution). This is because
at small residence time, due to diffusional limitations, the reactant sees only a
5.1 Membrane Reactors with Nonuniform Catalyst Distribution 99
Figure 5.4. A comparison of well-mixed IMRCF, CMR, and FBR with uniform [a(s) = 1 or
a(ξ ) = 1] and Dirac-delta (at s̄ = 1 or ξ̄ = 1) catalyst activity distributions for (a) product
purity on the feed side ζF,B , (b) product purity on the permeate side ζP,B , (c) product
molar flow rate on the feed side "F ζF,B , and (d) product molar flowrate on the permeate
side "P ζP,B , as a function of dimensionless residence time . The system is an isothermal
first-order reversible reaction A B. (From Yeung et al., 1994.)
small amount of the total catalyst in the pellets of the IMRCF, but it is exposed
to the full amount of catalyst in the FBR, thus resulting in higher conversion to
product.
Similar arguments hold when product purity or product flow rate on either the
feed or the permeate side is used as the performance index. The purities on the
feed and permeate sides are the product mole fractions ζF,B and ζP,B respectively,
while the product flow rates are "F ζF,B and "P ζP,B . A comparison for these cases
is illustrated in Figure 5.4. It can be seen (Figure 5.4.a) that on the feed side
both FBRs perform better than the membrane reactors for most cases. This is
because the feed side is depleted of the products as a result of diffusion across
the membrane. This point is illustrated in Figure 5.4.c, where a maximum may be
observed. For small residence times, the product molar flow rate is small because
of insufficient contact with the catalyst, whereas for larger residence times the
product permeates through the membrane until total permeation occurs (i.e. no
flow out from the feed side). When the product purity or its flow rate on the
permeate side is to be maximized, it is seen in Figure 5.4.b and d that for both the
100 Membrane Reactors
CMR and IMRCF a Dirac-delta distribution of catalyst located on the feed side
(i.e. ξ̄ = 1 for the CMR and s̄ = 1 for the IMRCF) shows better performance than
a uniform distribution.
The particular case of a planar catalytic membrane where the downstream bulk
flow is very large has been considered by Keller et al. (1984). In this case we can
assume that the downstream concentration of the reactant diffusing through the
membrane is zero. Under isothermal conditions and in the absence of external
transport resistances, the problem reduces to the diffusion–reaction equation in
the catalytic membrane:
d2 u
= φ 2 a(s) f (u) (5.4)
ds 2
together with the boundary conditions
s = 0: u=1 (5.5a)
s = 1: u = 0. (5.5b)
Note that this problem differs from that of a catalyst pellet, given by equations
(2.63)–(2.66), only in the BCs. The space coordinate of the membrane in this
case starts at the side in contact with the feed stream (s = 0) and finishes at the
downstream side (s = 1).
For reaction rates f (u) that increase monotonically with reactant concentra-
tion, and with effectiveness factor as the performance index, a Dirac-delta function
located at the feed side surface of the membrane is found to be the optimal catalyst
distribution, in accordance with the analysis in section 2.1. This is where the high-
est reactant concentration is encountered, which is beneficial for a positive-order
reaction.
A different answer for the optimization problem is obtained when minimizing
the flux of unreacted material leaving the membrane,
du
I= − . (5.6)
ds s=1
Using a specific variational analysis applied to the case of linear kinetics, i.e. f (u) =
u, Keller et al. (1984) found that the optimal activity distribution is given by the
following step function:
1/L, s2 ≤ s ≤ s2 + L
a(s) = (5.7)
0, 0 ≤ s < s2 and s2 + L < s ≤ 1
where
√
L
s2 = 1 − L − (5.8a)
φ
1 − 1 + φ2
L= 1+2 . (5.8b)
φ2
Even though both the objective functional (5.6) and the BCs (5.5) are not the
same as in the case of the catalyst pellet, it is surprising that here we encounter for
the first time an optimal activity distribution which is not a Dirac-delta function.
In order to understand this point, following the procedure described in section
102 Membrane Reactors
which is the same as (2.81) with n = 0, except that the function G ∗ is now given by
G ∗ = φ 2 λu ∗ (5.10)
and the Lagrange multiplier λ satisfies
d2 λ
= φ2a∗λ (5.11)
ds 2
with BCs
s = 0: λ=0 (5.12a)
s = 1: λ = 1. (5.12b)
By taking (5.7) as the optimal distribution a ∗ and substituting in equations (5.10)–
(5.12), we obtain
√
φ 2 s(1 − s − L)(1 − L)−2 exp(−φ L), 0 ≤ s < s2
√
∗ −1
G = φ [1 − s2 (1 − L) ] exp(−φ L),
2 2
s2 ≤ s ≤ s2 + L . (5.13)
φ 2 (1 − s)(s − L)(1 − L)−2 exp(−φ √ L), s + L < s ≤ 1
2
Vycor glass membrane containing 0.25% Pd, which was located preferentially
at the inner and outer surfaces of the glass tube (PM1), and (c) a Vycor glass
membrane containing 0.5% Pd, which was distributed uniformly throughout the
membrane wall (PM2). In all cases, total cyclohexane conversion increased with
increasing sweep-gas flow rate on the permeate side, due to selective and con-
tinuous removal of hydrogen from the reaction zone. In general, the presence of
palladium particles in the membrane decreased cyclohexane conversion owing to
the reverse reaction, and this effect was larger for PM2 than for PM1.
Yeung et al. (1996) studied ethane dehydrogenation to ethylene in an IMRCF
using Pt-Sn/γ -Al2 O3 catalysts. Three types of pellets were prepared with the same
total metal loadings (1.1 wt% Pt, 1.3 wt% Sn): (a) uniform Pt distribution, (b) nar-
row surface-step Pt distribution (Pt layer thickness ∼10% of pellet radius), and
(c) wide surface-step Pt distribution (Pt layer thickness ∼40% of pellet radius).
In all cases, Sn (added to enhance ethylene selectivity) was distributed uniformly
in the pellets. The membrane was a thin dense film of palladium deposited on
a porous 316 stainless-steel tube that permits transport only of hydrogen. As
shown in Figure 5.5, ethane conversion increases with residence time for both
nonuniform distributions, and is two to three times higher than the equilibrium
conversion. The highest conversion is attained by concentrating the Pt catalyst
at the pellet external surface. This indicates that the influence of transport pro-
cesses in the pellet is not negligible and their presence is detrimental to catalyst
performance.
Szegner et al. (1997) studied experimentally and theoretically the same reac-
tion on step-type Pt-Sn/γ -Al2 O3 catalysts in a well-mixed isothermal IMRCF.
Four types of pellets with different Pt distributions but with approximately the
same amount of platinum and tin were used, as shown in Figure 5.6. The membrane
used was a composite mesoporous γ -Al2 O3 membrane. In order to minimize reac-
tant loss, the effective permeability of the commercial composite MembraloxTM
membrane with 2-µm γ -Al2 O3 membrane layer was reduced by slip-casting of
alumina sol. The slip-cast modified membrane was approximately 17 µm thick
(see Figure 5.7 for cross sections of original and slip-cast modified Membralox
membranes). As shown in Figure 5.8, optimum reactor performance was ob-
tained when the catalyst was concentrated at the surface of the pellet. In this case
104 Membrane Reactors
results was good (see Figure 5.8), especially considering that no adjustable param-
eters were used, since intrinsic reaction kinetics and diffusivities were determined
from separate experiments.
where Kg is the permeability of the gel layer, Pg is the pressure drop across the
Figure 5.11. Membrane layer thickness as a function of slip-cast parameters for inert
2-wt% Al2 O3 and active 5-wt% Pt/alumina –2-wt% Al2 O3 sols. (a) Parameter: (time)1/2 ;
inert and active A1 sol. (b) Parameter: number of slip-cast layers; inert and active A1 sol,
(c) parameter: (time)1/2 ; inert and active A2 sol. (From Yeung et al., 1997.)
gel layer, t is time, η is the solvent viscosity, and αg is a constant dependent on both
the sol concentration and the support characteristics. Besides increasing the slip-
casting time, the thickness of the membrane can also be increased by multiple
slip-casting, as shown in Figure 5.11.b. In this experiment, it is observed that the
inert and active sols display different behavior after the first slip-cast, because
the support on which subsequent layers are deposited is different in the two cases.
In Figure 5.11.c, the dependence of layer thickness on time is shown for inert (A2)
and active (Pt-A2) sols. The A2 sols exhibit different behavior from the A1 sols
(compare Figure 5.11.a and c). For the same slip-casting time, sol A1, which has
larger alumina particles, gives a thicker layer than A2. However, the active layer
Pt-A2 is thicker than Pt-A1.
Figure 5.12.a shows the dependence of layer thickness (slip-casting time 5 s)
on the alumina concentration of the slip. The thicknesses for both A1 and A2
sols increase with alumina concentration, reaching a maximum at about 15 wt%.
The viscosity of the slip also increases with alumina content as shown in Figure
5.12.b. Both sols display sharp increases in viscosity for alumina concentration
5.4 Experimental Studies 109
above 10 wt%, which results in the thickness leveling off as observed in Figure
5.12.a. Sol viscosity is important in preventing pore clogging as well as controlling
gelation rate of the slip-cast (Cot et al., 1988). The viscosity can also be controlled
by addition of binders such as PVA, which help regulate the gelling rate and reduce
crack formation in the final membrane.
6
Special Topics of Commercial Importance
I
n this chapter, we consider specific topics of significant commercial value where
nonuniform distribution plays an important role in catalyst design. These in-
clude catalysts for automotive exhaust cleanup, petroleum refining operations
such as hydrotreating and cracking, biotechnology, and acid catalysis. Particularly
in the case of automotive catalysis, nonuniform distribution of noble metals pro-
vides critical advantages for pollution abatement reactions and has been employed
extensively.
110
6.1 Automotive Exhaust Catalysts 111
its strong sensitivity to phosphorus and lead poisoning, suggests using it in small
quantities with protection from poisoning. Based on these considerations, in each
pellet the noble metals were located in three separate layers, one underneath the
other, as illustrated in Figure 6.3. The first, at the pellet’s external surface, was
the Pt layer whose depth was determined so as to entirely contain the expected
penetration depth of the poison during the desired catalyst lifetime. Next was
located the Rh layer, so that diffusional resistances for NOx reduction would be
minimized and Rh would also be protected from poisoning. Finally, the Pd layer
was placed at the deepest location, providing good lightoff performance, especially
for aged catalysts, because of its good thermal stability. This configuration also
prevented undesirable alloy formation among the noble metals. While maintaining
the same noble metals and their order in the layer configuration, Vayenas and
coworkers (Vayenas et al., 1994; Papadakis et al., 1996) have proposed the use
of different supports that further enhance catalytic activity. Multilayer three-way
catalyst design remains an active area of research, and numerous publications
continue to appear in the patent and scientific literature.
There is a strong motivation to avoid using Rh due to its high cost, and hence
research has been directed towards alternative catalyst formulations. However,
Pd- and Pt-based catalysts are typically not as effective for NO conversion. Some
researchers have tried to overcome these disadvantages by using various additives.
Hu et al. (1996) improved the performance of Pd-only three-way catalysts using
a two-layer design, where the first layer contained Pd and the second Pd and
ceria. Both layers utilized alumina as the support and contained base-metal oxide
stabilizers. The Pd layer provided high activity at low temperatures, while the Pd–
Ce layer offered large oxygen storage capacity, which gave improved three-way
activity at high temperatures. The two-layer Pd catalyst exhibited performance
comparable to a Pt/Rh catalyst.
Automotive exhaust catalysts constitute a remarkable example of catalyst de-
sign despite the fact that in many instances, because of complexity of the problem
and lack of detailed expressions for the kinetics of the various reactions involved,
the design was based on experimental observations rather than on quantitative
models.
concentration is low (Oh and Cavendish, 1983). Psyllos and Philippopoulos (1993)
numerically studied the performance of monoliths with various parabolic axial
catalyst distributions, and showed that certain distributions have shorter warm-up
periods than the uniform one. Tronci et al. (1999) demonstrated theoretically that
further improvements can be obtained by optimizing the axial catalyst distribu-
tion. By using the optimal distribution, lightoff time was decreased significantly as
compared to the uniform distribution. A two-zone converter was also investigated,
and resulted in similar performance to that of the optimal distribution.
A two-dimensional catalyst distribution pattern, both axially along and radially
across the monolith, was investigated by Naoki et al. (1996), and is shown in Figure
6.5. The quantity of noble metal was six times the base loading in the outer region
of the first zone, equal to the base loading in the inner region of the first zone,
and one-half the base loading in the second zone. This catalyst distribution was
shown theoretically to exhibit higher conversion efficiency during cold start than
did a uniform monolith with the same total catalyst amount. The investigators
considered radial variations in inlet velocity, inlet temperature, radial heat transfer,
and diabatic operation. High loading in the first zone was found to improve cold-
start behavior, in agreement with the discussion above. Concentrating the catalyst
towards the central region of the first zone improved conversion at the early stage
of warm-up, while concentrating it towards the periphery, as shown in Figure 6.5,
increased conversion at later stages of the warm-up, after lightoff of the central
region. The latter design exhibited the best performance and could be improved
further by insulating the monolith.
deposition. No restrictions were placed on either the local or the overall catalyst
loadings in seeking an optimal solution that maximizes the amount of contam-
inant metal deposition over the useful catalyst life. The optimal catalyst distri-
bution was calculated numerically. The performance of a nonuniform catalyst
was compared with that of a uniform catalyst which was already optimized for
catalyst loading and pore diameter. As expected, the obtained optimal distribu-
tion exhibited a sharp maximum towards the center of the pellet followed by
a decrease towards the external surface. This distribution provided a substan-
tial increase in the overall amount of metals deposited, as shown in Figure 6.7.
As approximations of the optimal distribution, edge-reduced (where the catalyst
distribution was uniform, but reduced only at the outer edge of the pellet) and
egg-yolk catalysts were also considered. It was found that the total metal deposited
in a uniform pellet could be increased by nearly 25% utilizing the edge-reduced
profile.
Chiang and Tiou (1992) extended the above ideas to the fixed-bed reactor. In
order to explore the effects of nonuniform catalyst distribution, they considered
three specific cases: linear (increasing towards the center), edge-reduced, and egg-
yolk catalysts, with the same total loading. The evaluation was based on the reactor
metal deposition capacity at the point where the metal compound concentration in
the reactor effluent reached 30% of the inlet value. The linear profile exhibited the
largest improvement in metal deposit (almost 50%) over the uniform distribution.
The edge-reduced and egg-yolk catalysts also showed significant improvements
118 Special Topics of Commercial Importance
over the uniform pellets. For each catalyst distribution, the initial pore size of the
pellet was optimized. The optimal pore sizes of the nonuniform distributions were
smaller than those of the uniform one, because they increased the pore surface area
for metal deposition while simultaneously delaying the plugging process. Further
improvement was realized with two- and three-zone reactors, but the additional
improvement was smaller for the nonuniform distributions.
The case of a moving-bed reactor was also investigated (Chiang and Fang,
1994), using two families of active catalyst distributions: linear and two-step. Pore
size and active catalyst distribution were optimized simultaneously. The benefit
of using nonuniform distributions over the uniform was small, in contrast to the
fixed bed reactors discussed above, and can be attributed to the short residence
time of catalyst in the moving-bed reactor. The optimal pore size and total metal
deposit were both larger for the fixed bed than for the moving bed.
In all the above studies pore connectivity was not taken into account. Arbabi
and Sahimi (1991a,b) developed a percolation model which predicts the occurrence
of a transition point (the percolation threshold) where the previously connected
cluster of unplugged pores becomes disconnected, so that macroscopic transport
and reaction are no longer possible. Similar catalyst performance improvements
obtained by Limbach and Wei (1988) for nonuniform catalyst distribution were
also observed. Nonuniform catalyst distributions with activity increasing towards
the pellet center lead to increased metal deposition capacities.
For hydrodesulfurization (HDS), Asua and Delmon (1987) investigated the-
oretically the behavior of catalysts containing molybdenum and cobalt sulfides.
In this case, catalyst performance optimization can be pursued by varying both
the overall catalyst concentration (i.e. Mo+Co) as well as its composition (i.e.
Mo/Co). They used a detailed kinetic model which accounted explicitly for the
catalyst composition, as well as for the concentration of active sites, and showed
that the catalyst performance could indeed be improved by using nonuniformly
distributed catalyst pellets.
Goula et al. (1992a) studied experimentally the hydrodesulfurization of thio-
phene on eggshell, egg-white, egg-yolk, and uniform MoO3 /Al2 O3 catalysts. All
catalysts had similar loadings (12–14% MoO3 ) and dispersions (0.098–0.134), but
different numbers of active sites. The highest activity for HDS was found for
the egg-white catalysts, and it was related to the number and type of active sites
rather than the Mo distribution. However, from powder experiments it was es-
tablished that diffusional resistances were present, and the effectiveness factor
followed the trend η eggshell > η egg-white > η egg-yolk . On the other hand, the selec-
tivity towards butane, which is formed by hydrogenation of the unsaturated hy-
drocarbons produced by HDS, was affected by the catalyst distribution and fol-
lowed the trend Segg-yolk > Segg-white > Seggshell . This behavior was related to the
residence time of unsaturated hydrocarbons inside the pellets. In addition, the
uniform catalyst exhibited the lowest activities for both hydrodesulfurization and
hydrogenation, and this was attributed to partial clogging of the pores during
preparation.
6.3 Composite Zeolite Catalysts 119
A→B→C (6.2)
→
→
When the intermediate B is included in the reaction set, then for the case where
the rate constants of all reactions but B→C are greater in the silica–alumina
matrix than in the zeolite, the diluted egg-yolk distribution leads to higher yield
and selectivity than the uniform distribution. This occurs because the formation
of component B in the outer region can lead to a greater amount of desired C in
the inner region. The optimum radius of the diluted core and the optimum yield
both depend on the values of the parameters. When the overall volume fraction of
silica–alumina was also considered as an optimization variable, it was found that
yield exhibited a maximum with respect to both the diluted-core radius and the
volume fraction of silica–alumina.
Dadyburjor and coworkers (Martin et al., 1987; White and Dadyburjor, 1989;
Dadyburjor and White, 1990) studied extensions of the above schemes including
changes in reaction orders as well as position-dependent rate constants and diffu-
sivities arising from nonuniform poisoning. Since these variations induce complex
interactions between transport and reaction in the zeolite and matrix, a variety
of zeolite distributions were found to be optimal, depending on the specific case
investigated.
One of the early studies was by Horvath and Engasser (1973), who investigated
theoretically the properties of pellicular heterogeneous catalysts, which consist of
a fluid-impervious core supporting a spherical annulus of the catalytically active
porous medium. This configuration is equivalent to an eggshell enzyme distribu-
tion. For Michaelis–Menten kinetics under isothermal conditions, it was shown
that a fixed amount of enzyme gives higher overall reaction rate when it is con-
fined to the outer shell than when it is distributed uniformly in a porous pellet of
the same diameter. As expected, significant improvements in overall reaction rate
are realized by the eggshell catalyst for high Thiele modulus and sufficiently low
reactant concentration. The latter occurs at the end of batch operation, when high
reactant conversions are required. These conclusions were confirmed experimen-
tally for eggshell and uniformly distributed agarose–micrococcal nuclease in the
hydrolysis of a phosphate compound, which follows Michaelis–Menten kinetics
(Guisan et al., 1987). Under conditions of low reactant concentration and high
Thiele modulus, the highest increase of effectiveness factor obtained theoretically
was 35%, while a 16% improvement was observed experimentally.
Higher effectiveness factors for eggshell immobilized trypsin on derivatized
glass carriers, as compared to a uniform distribution, were demonstrated experi-
mentally and theoretically by Buchholz (1979) and Borchert and Buchholz (1979,
1984). For the hydrolysis of low concentration N-α-benzoyl-L-arginine ethyl ester
solutions, experimental effectiveness factors for eggshell distributions were higher
by up to 50%; when the higher-molecular-weight reactant casein was used, the
effectiveness factor was higher by up to 100%, and this was attributed to large in-
traparticle concentration gradients. The improvements in effectiveness were more
significant when larger support particles were used, since the Thiele modulus was
then higher.
Because of the sensitivity of enzyme activity to pH, another reason for infe-
rior performance of the uniform distribution could be slow proton diffusion from
inner core of the particle to solution. Carleysmith et al. (1980a) showed that for
deacylation of benzylpenicillin catalyzed by penicillin acylase, slow proton diffu-
sion resulted in low reaction rates even when reactant diffusional limitations were
decreased by the use of small particles. It was postulated that a pH gradient ex-
isted in the beads even in the presence of a buffer. On the other hand, for eggshell
distributions, all the enzyme probably operated near its optimum pH.
Subsurface enzyme distributions can confer a measure of protection against
degradation by local concentrations of harsh reagents which may need to be added
during reaction. Carleysmith et al. (1980b) showed experimentally for the hydrol-
ysis of benzylpenicillin that when the enzyme (penicillin acylase) was immobilized
in a subsurface layer, the particles retained their initial activity to a larger extent
after three consecutive batch reactions than with a shell-type enzyme distribution.
This was attributed to the fact that the enzyme was protected from localized re-
gions of high pH that resulted when alkali was added to maintain the pH of the
reaction medium constant.
Park et al. (1981) extended Horvath and Engasser’s work by studying theo-
retically the behavior of eggshell, egg-yolk, and uniform enzyme distributions for
6.4 Immobilized Biocatalysts 123
show higher conversions than decreasing distributions at large operating times in-
dicates that the optimal enzyme concentration is an appropriate subsurface Dirac
distribution.
Chung and Chang (1986) compared theoretically the performance of nonuni-
form biocatalysts for first-order kinetics in the presence of two different deactiva-
tion mechanisms representing the effects of temperature and time. The first was
based on a reversible thermal denaturation model, according to which enzyme
activity was an exponentially decreasing function of temperature (Ollis, 1972).
This, combined with the usual Arrhenius dependence, gives rise to a maximum
in the first-order reaction rate constant as a function of temperature. For the sec-
ond deactivation mechanism, the reaction rate constant was assumed to decay
exponentially with time. Three distributions with the same amount of enzyme
were considered: uniform, eggshell, and egg-yolk. Egg-yolk immobilized enzymes
had the smallest effectiveness factors, but they showed the best stability under
temperature fluctuations. The egg-yolk distribution also exhibited better stability
than eggshell catalysts over time, because in the former the effect of rate-constant
decrease is counterbalanced by better utilization of the enzyme arising from the
concomitant decrease of the Thiele modulus. Thus, stability with time increased
in the order eggshell to uniform to egg-yolk.
Distribution
Macroparticle Microparticle Synthesis Type Figure
H2 SO4
completely A
sulfonated
was much slower than in the macropores. Therefore a uniform profile was obtained
in the macroparticle, while the microparticles exhibited a surface distribution. For
particles of type C, using C2 H4 Cl2 as swelling agent, decreasing profile towards the
center was obtained at both scales. Finally, for particles of type D, where the most
active sulfonating agent SO3 was used (in a gas-phase fluidized bed with nitrogen)
together with a relatively short sulfonation time, the sulfonic group concentration
profile was steeper in the macropores than in the microparticles.
Verification of the distributions discussed above was obtained by measuring
the concentration of sulfonic groups by energy-dispersive X-ray analysis (EDX)
of cross-sectional areas of resin particles (Bothe, 1982; Klein et al., 1984). The
available resolution permitted measurement of profiles only for the macroparti-
cles, not for the microparticles. In Figure 6.8 sulfonic-group concentration profiles
measured as a function of time during sulfonation of type B particles are shown.
The profile in the macropores is indeed uniform, and concentration increases with
time as expected. Although the microparticle profiles cannot be measured, results
reported by Ahn et al. (1988) under similar conditions show that gelular resins
are not swollen and that sulfonation occurs only at the external surface. Finally,
Figure 6.9 confirms the sulfonic groups’ concentration profiles indicated earlier
for type C and D particles, respectively (see Table 6.1).
used the alkylation of benzene with propylene as a model reaction. Due to the
nonpolar nature of the reactants, the gelular microparticles were not swollen, and
therefore the acid groups located inside the microparticles were not accessible
to the reacting liquid. This explains the higher activity observed for the catalyst
of type C in Table 6.1, as compared to uniform particles A, as well as to particles
of type D, in which the largest fraction of acid groups are inside the gelular phase
and are hence inaccessible.
Ahn et al. (1988) considered the isomerization of 1-butene as a model reac-
tion to investigate the performance of three types of macroporous resin particles.
Two of these, obtained by sulfonation of preswollen resin particles, were uniform
but with different total loading. The third was of type B in Table 6.1, i.e., it con-
tained sulfonic groups only on the surface of the microparticles, but uniformly
distributed throughout the entire macroparticle. The results obtained were inter-
preted through a two-phase model, where the simultaneous diffusion and reac-
tion processes were described at both the microparticle and macroparticle levels.
In particular, the overall effectiveness factor of the macroporous particle, η was
computed from
η = ηα [β + (1 − β)ηι ] (6.6)
where ηα and ηι are effectiveness factors accounting for diffusion in the macrop-
ores and the microparticle, respectively, while β is the fraction of sulfonic groups
on the external surface of the microparticles, which are accessible immediately
through diffusion in the macropores. The experiments were found to be consis-
tent with the model results, leading to values of the microparticle effectiveness
factor in the range 0.1–0.3 and to a reaction order with respect to sulfonic-group
concentration equal to 2.4. The latter was explained by considering that the rate-
determining step of the isomerization involved 2–3 sulfonic groups, in agreement
with the results reported by Gates et al. (1972) for t-butanol dehydration. Catalyst
B showed the best performance, since for a given loading of acid groups the local
concentrations were higher and the reaction was faster.
Chee and Ihm (1986) investigated the effect of nonuniform distribution of ac-
tive sites on catalyst deactivation during the gas-phase ethanol dehydration. In
this system the produced water molecules interact with pairs of sulfonic groups,
either forming hydrogen bonds or reacting with them to cause resin desulfonation.
Both mechanisms were considered in a deactivation kinetic expression which was
second-order with respect to sulfonic-group concentration. The two-phase model
mentioned above was used to simulate the deactivation process observed exper-
imentally in two types of sulfonated membranes. The macroporous membranes
utilized had the same total loading of sulfonic groups but different distributions:
one was uniform, while in the other only the external surface of the microparticles
was functionalized, i.e. type A and B in Table 6.1, respectively. The observed de-
activation patterns were quite different, and the nonuniform membrane exhibited
the fastest deactivation rate due to the high surface concentration of acid groups.
The experimental observations were in qualitative agreement with the two-phase
model.
6.5 Functionalized Polymer Resins 129
which produces isobutene and methanol. Undesired reactions are mainly metha-
nol dehydration and isobutene dimerization,
which are both catalyzed by acids. When a uniform resin is employed, the nonpolar
ether reactant does not enter the sulfonated gelular microparticles, and the main
reaction is restricted to the surface active groups in the macropores. However, the
consecutive reactions, and particularly reaction (6.8) involving a polar reactant,
take place also in the microparticles. Accordingly, when using a resin sulfonated
only on the external surface of the microparticles, i.e. type B in Table 6.1, the
product of the main reaction diffuses out of the macropores, reaction (6.8) is prac-
tically eliminated, and nearly 100% selectivity to the desired product is achieved
(Widdecke et al., 1986).
Similar selectivity issues are pertinent to esterification and transesterification re-
actions, where an ester is produced by reacting an alcohol with an acid or another
ester, respectively. These are large families of reactions, relevant to various prod-
ucts of industrial interest (Lundquist, 1995). In order to achieve satisfactory reac-
tion rates, high temperatures (60–120◦ C) must be employed, which unfortunately
give rise to undesired reactions catalyzed by the sulfonic groups. As mentioned
above, these include the production of undesired ethers by alcohol condensation.
Additionally, in the case of secondary or tertiary alcohols, dehydration reactions
may also occur, leading to olefinic by-products. When uniform resins are utilized,
the polar alcohol molecules have a strong tendency to enter the sulfonated gelular
microparticles, where both the above undesired reactions can take place. Hence,
similarly to ether cleavage, selectivity is improved by confining the sulfonic groups
to the external surface of the microparticles. There they are equally accessible by
the alcohol as well as the other reactant, i.e. acid or ester for esterification and
transesterification, respectively.
Gelosa et al. (1998) studied the esterification of phthalic anhydride with ethanol,
using type A and B resins (see Table 6.1). The reaction system consists of the two
consecutive reactions
types of sulfonated resin particles, i.e. uniform (type A in Table 6.1) and surface
distribution (type B), respectively. The conversion profiles are similar as shown
in Figure 6.10, but show significant differences in selectivity. In particular, for the
homogeneous catalyst no by-products are formed, since the alcohol is contacted
by the acid catalyst only in the presence of the monoethylphthalate, in which
case only the esterification reaction is favored. However, when the uniform resin
is utilized, ethanol accumulates in the gelular sulfonated microparticles, where
the monoethylphthalate is almost completely excluded. This favors the ethanol
condensation reaction, which leads to ∼2.5% ether formation at the end of the
reaction. This reaction is minimized with the surface-sulfonated resins, which lead
to only 0.4% ether formation.
7
Preparation of Pellets with Nonuniform
Distribution of Catalyst
N
onuniform catalyst distributions in porous supports are obtained primarily by
multicomponent impregnation techniques. In general, an intermediate level
of interaction between catalyst precursor and support is required, so that the
precursor can attach to the support, but can also desorb if another competing
adsorbing species is present. Depending upon the interplay between competitive
adsorption and diffusion of the various species in the porous support, a variety
of nonuniform catalyst distributions can be obtained. The above physicochemical
processes are also encountered in chromatographic separations (Ruthven, 1984).
This chapter is divided in two parts. The first deals with adsorption on powders,
while the second is focused on simultaneous diffusion and adsorption phenomena.
Although diffusion–adsorption methods are dominant for the preparation of
nonuniform catalyst pellets, other procedures have also been employed. One such
technique is deposition precipitation in preshaped carriers (De Jong, 1991). It in-
volves deposition inside pellets of insoluble compounds, such as hydroxides which
are formed by a precipitation reaction. The latter can be induced by a change of
solution pH. Immediately after imbibition, a pH profile develops inside the pellets,
which depends on the initial solution pH and the isoelectric point of the carrier.
Since precipitation reactions depend on pH, the insoluble compound distribution
reflects the pH gradient. Hence, by appropriate choice of the impregnation con-
ditions, precipitation can occur in either the inner (egg-yolk distribution) or the
outer (eggshell distribution) region of the pellet. However, preparation of eggshell
catalysts leads to the problem of precipitation outside the pellets. Therefore this
method is especially suited for preparation of egg-yolk distributions, and has been
utilized for preparation of Mo/SiO2 , Cu/γ -Al2 O3 , and Ag/γ -Al2 O3 catalysts.
In other preparation techniques, the nonuniform catalyst pellet is obtained by
assembling catalytic and noncatalytic layers. One such method is based on cata-
lyst powder granulation. This can be realized by a two-stage fluidized/spouted bed
(Scheuch et al., 1996). Catalyst powder premixed with a binder is deposited grad-
ually on moist seeds (typically 0.5–2-mm diameter) of catalyst support, ultimately
providing layered spherical particles. This method can in principle be extended for
preparation of egg-white catalysts by using the corresponding eggshell catalysts
131
132 Preparation of Pellets with Nonuniform Distribution of Catalyst
as seed granules and depositing inert powder. Similarly, egg-yolk catalysts could
be produced using uniform spherical catalyst particles as seed granules.
Coating is another technique that has been used for the preparation of egg-
shell catalysts (Pernicone and Traina, 1984; Stiles and Koch, 1995). The coating
equipment generally resembles pill-coating devices used in the pharmaceutical
industry. The cores are coated in a rotating drum, with an appropriate catalytic
paint. The thickness of the layer can be controlled by the amount of paint slurry
introduced in the drum.
Tableting can also in principle be used to prepare egg-shell catalysts or other
nonuniform distributions. In fact, specially designed presses are used in the phar-
maceutical industry to prepare multilayer tablets. Examples of nonuniform pellet
preparation by pressing together different layers are reported by Wu et al. (1990b)
and Gavriilidis and Varma (1992).
Finally, extrusion can also be adapted for the preparation of nonuniform extru-
dates by appropriate design of the die.
be calculated by setting the transient term equal to zero, thus giving the Langmuir
adsorption isotherm
ne KCe
= (7.2)
ns 1 + KCe
where ne is the adsorbed amount, and Ce is the solution concentration, both at
equilibrium. K = k+ /k− is the equilibrium adsorption constant, and a large value
indicates strong adsorption. The Langmuir isotherm was first used to describe
adsorption of gases on solids, but was later extended to describe adsorption from
solutions. It assumes that the surface is energetically uniform, the energy of ad-
sorption is independent of surface coverage, and the solution is dilute, so that the
adsorbed species do not interact with each other. This simple equation represents
well a broad spectrum of experimental adsorption data.
Adsorption of many metal precursors on various supports has been studied
under equilibrium and transient conditions in batch systems. The most widely in-
vestigated system is H2 PtCl6 on γ -Al2 O3 . A Langmuir model of adsorption was
generally found to be suitable for fitting the experimental data (Santacesaria et al.,
1977a; Shyr and Ernst, 1980; Castro et al., 1983; Jianguo et al., 1983; Heise and
Schwarz, 1988; Xidong et al., 1988; Subramanian and Schwarz, 1991; Blachou and
Philippopoulos, 1993; Papageorgiou et al., 1996). Figure 7.1 shows experimental
results of H2 PtCl6 adsorption on γ -Al2 O3 under (a) equilibrium and (b) transient
conditions. The surface saturation coverage ns and adsorption equilibrium con-
stant K can be determined from equilibrium measurements, and the adsorption
rate constant k+ from transient measurements. The surface saturation coverage
obtained in Figure 7.1.a is ns = 1.55 µmol/m2 , and the equilibrium adsorption
Area K k+ ns te
(m2 /g) (l/mol) (l/mol · s) (µmol/g) (µmol/m2 ) (h) Reference
177 459 1.21 265 1.50 3–8 Santacesaria et al., 1977a
150 1,330 110 0.73 2 Shyr and Ernst, 1980
170 31,000 140 0.82 6 Castro et al., 1983
245 275 1.12 6 Jianguo et al., 1983a
7,200 1.70 Scelza et al., 1986
190 1,550 151 0.79 0.3 Heise and Schwarz, 1988
195 110 0.18 290 1.48 2 Subramanian and Schwarz, 1991
227 205 0.90 1 Mang et al., 1993
78 606 2.46 121 1.55 1 Papageorgiou et al., 1996
a Solid was η-Al2 O3
constant is 606 l/mol. From Figure 7.1.b, the adsorption rate coefficient k+ was cal-
culated to be 2.46 l/mol · s. Values of these parameters reported in the literature
range for surface saturation coverages from 0.73 to 1.55 µmol/m2 , and for the ad-
sorption equilibrium constant from 110 to 31,000 l/mol (see Table 7.1). Adsorption
characteristics of other systems that have been investigated include compounds
of Pd on alumina (Sivaraj et al., 1991), Pd on silica and alumina (Schwarz et al.,
1992), Pt on carbon (Hanika et al., 1982, 1983; Machek et al., 1983a,b), Rh on
alumina (Hepburn et al., 1991a), Ni on alumina (Komiyama et al., 1980; Huang
et al., 1986; Huang and Schwarz, 1987; Clause et al., 1992), Ni, Ba on alumina –
one- and two-component impregnation – (Melo et al., 1980b), Cr on alumina
(Chen and Anderson, 1973), Mo on alumina (Wang and Hall, 1980), Cr, Cu on
alumina – one- and two-component impregnation – (Chen and Anderson, 1976),
Cr, Mo, W on alumina – one-component impregnation – (Iannibello et al., 1979),
Co, Mo on alumina – one- and two-component impregnation – (Iannibello and
Mitchell, 1979; Cheng and Pereira, 1987; Hanika et al., 1987a,b; Sporka and
Hanika, 1992), and organic acids on alumina (Kummert and Stumm, 1980; Engels
et al., 1987).
The adsorption capacity of a support can be altered by calcination prior to im-
pregnation (Schwarz, 1992). The number of surface sites on alumina which can
be used for binding the metal precursor changes in a nonmonotonic fashion as a
function of calcination temperature. In particular, a minimum in acidity and sur-
face sites has been observed at about 600◦ C by some investigators (Tanabe, 1970;
Sivaraj et al., 1991). Santacesaria et al. (1977b) found that the amount of plat-
inum adsorbed during hexachloroplatinic acid impregnation on alumina samples
calcined at different temperatures followed the same trend as the acidity.
In addition to the Langmuir isotherm, other models such as the Freundlich
isotherm have been used for fitting adsorption data (cf. Benjamin and Leckie,
1981; Haworth, 1990). This isotherm is described by
ne = KCe1/m (7.3)
7.1 Adsorption on Powders 135
Contescu and Vass (1987) studied the adsorption of chloro- and aminopalla-
dium complexes on alumina and found that at constant pH the adsorption data
followed the Langmuir isotherm. Both the adsorption capacity ns and the adsorp-
tion constant K changed with pH and showed minima at the isoelectric point of
alumina. For molybdate adsorption on alumina, it has been demonstrated that
the adsorption sites are protonated hydroxyls, whose concentration can be de-
creased by increasing the pH (Spanos et al., 1990a,b; Goula et al., 1992b). Huang
(1975) studied the adsorption of ortho-phosphate on hydrous γ -Al2 O3 from dilute
aqueous solution at constant ionic strength. Adsorption followed the Langmuir
isotherm, except for large values of pH (>10). The adsorption constant K was
found to vary with pH, and the adsorption capacity ns showed a maximum at
pH = 4, decreasing rapidly with increasing pH.
In all systems above, a decrease in anion adsorption was observed at pH larger
than a certain value. This can be explained by invoking the simple electrostatic
model of Brunelle (1978). According to this model, when oxide particles are sus-
pended in aqueous solutions, a surface polarization results in net electrical surface
charge. The type and magnitude of this charge are a function of the pH of the so-
lution surrounding the particle and the nature of its surface. In general, in acidic
solutions the surface is positively charged (S–OH+ 2 ), while in basic solutions the
−
particles carry a negative surface charge (S–O ). This is conveniently expressed
by the following equilibrium reactions (Parfitt, 1976):
S–OH+ % +
2 & S–OH + Hs (Ka1 ) (7.4a)
← Decreasing pH
& S–O− + H+
S–OH % s (Ka2 ). (7.4b)
Increasing pH →
In between these two cases, a pH exists at which the net charge of the surface is zero.
This value is characteristic of the oxide, and is called point of zero charge (PZC).
Some authors use the terms point of zero charge (PZC) and isoelectric point (IEP)
interchangeably. Properly, PZC is associated with a zero overall charge of the sur-
face determined by potentiometric titration, and IEP with a zero electrophoretic
mobility determined by microelectrophoresis (Parfitt, 1976). When the PZC and
IEP points coincide (i.e. pHPZC = pHIEP ), the pHPZC is related to the two intrinsic
acidity constants (7.4) by the following equation (Hohl and Stumm, 1976; Schwarz
et al., 1992; Zhukov, 1996):
pKa1 + pKa2
pHPZC = . (7.5)
2
The protonated or deprotonated surface hydroxyl groups [see (7.4)] can be used to
fix the metal ions to the support surface. In acidic solutions the positively charged
surface will preferentially adsorb anions, while in basic solutions the negatively
charged surface will adsorb cations (Brunelle, 1978; D’Aniello, 1981; Foger, 1984;
Spanos et al., 1991). For example, for anionic adsorption the mechanism would be:
+
S–OH + H+ %
& S–OH2 (7.6a)
y(S–OH+
2)+M
−n % +
& (S–OH2 ) y M−n . (7.6b)
7.1 Adsorption on Powders 137
In the absence of chemical reactions, the amounts of anion adsorbed can be con-
trolled by adjusting the pH of the impregnating solution. Equivalently, a similar
mechanism can be postulated for cationic adsorption in a basic solution:
of ions which adsorb with one or two hydration sheaths intact, whereas Contescu
et al. (1993a, 1995a) attributed it to steric constraints, which force the geometry
of adsorbing complexes to match the arrangement of charged surface sites.
Mn –OHnv %
2 & Mn –OH
nv−1
+ H+ (Kn,1 ) (7.8a)
Mn –OHnv−1 %
& Mn –Onv−2 + H+ (Kn,2 ) (7.8b)
where n = 1, 2, 3 is the number of metal cations coordinated with surface –OH, and
v is the bond valence reaching the oxygen or hydroxyl. If one assumes the presence
of only one type of reactive surface group and if nv equals 1, the reactions (7.8)
simplify to the reactions (7.4). The intrinsic affinity constants Kn,i of the various
surface groups depend on the local configuration of the surface, i.e. the number of
coordinating cations, the valence of the cation, and the coordination number of
7.1 Adsorption on Powders 139
the central cations in the crystal structure. The difference between log Kn,1 and log
Kn,2 for proton binding on groups with the same surface configuration is so high
(about 14 log K units) that only one protonation reaction can take place for each
surface group within the normal pH range (3–11). Therefore, several Langmuir
isotherms can coexist, one for each surface site configuration, i.e. n = 1, 2, 3.
Schwarz and coworkers (Contescu et al., 1993a,b, 1995b) pictured the oxide
surface as consisting of oxo and hydroxo groups characterized by different proton
affinities. Using potentiometric titration data, they developed a method for calcu-
lating the proton affinity distribution of binding sites. In this way, the heterogeneity
of surface proton-binding sites was demonstrated. For the case of γ -Al2 O3 , it was
proposed that four categories of surface sites contribute to proton binding and sur-
face charge development between pH 3 and 11. They were identified with types I-a
(terminal, bound to tetrahedral Al), I-b (terminal, bound to octahedral Al), II-a
(bridging, bound to octahedral and tetrahedral Al), and III (triple-coordinated,
bound to octahedral Al) of surface hydroxyls. Type III are the most acidic sites,
while type I are the most basic ones. These surface groups react with solution
protons, depending on their log Ki and the solution pH, as follows:
[(Al Oh )3 –O]−0.5 + H+ %
& [(Al Oh )3 –OH]+0.5 , log KIII ∼ 2.5
[(Al Oh )–O–(AlTd )]−0.75 + H+ %
& [(Al Oh )–OH–(AlTd )]+0.25, log KII-a ∼ 4.1
−0.25 + %
[(AlTd )–OH] + H & [(AlTd )–OH2 ]+0.75 , log KI-a ∼ 6.8
[(Al Oh )–OH]−0.5 + H+ % & [(Al Oh )–OH2 ]+0.5 , log KI-b ∼ 9.8.
(7.9)
the pH results in variations in both number and type of surface sites with posi-
tive and negative charges. Depending on the solution pH, one or more types of
surface sites may contribute to adsorption of the catalyst precursor. Hence ad-
sorption equilibrium can be described by a combination of Langmuir isotherms,
each associated with one type of surface site contributing to adsorption (Contescu
et al., 1993a). This model predicts that even at the apparent PZC, some of the sites
are still charged. This explains adsorption data which contradict Brunelle’s model,
such as the adsorption of cations at pH < pHPZC , which in the case of γ -Al2 O3
is attributed to a negative charge of II-a sites at low pH (Contescu et al., 1995b).
In addition, geometrical constraints between the size and shape of the adsorbing
ions and the ensembles of charged surface sites required for adsorption may also
have to be allowed for, since these can be important factors that determine the
amounts of catalyst precursor adsorbed (Contescu et al., 1993a, 1995a).
sphere of the immobilized species is changed), and the formation of new chemical
compounds at the interface (the chemical identity of the immobilized species is
changed) (Westall, 1986; Schwarz et al., 1995). The interaction of metal precursor
with the support depends on its actual form in solution, i.e. its chemical specia-
tion. The latter is affected by the properties and concentration of the precursor,
the solution pH, the ionic strength, and the presence of added or extraneous
counterions (Schwarz, 1992). Metal ions in water are not bare, but hydrated. Thus
they can participate in exchange reactions where the coordinated water molecules
are exchanged for some preferred ligands. Hydrated multivalent metal ions can
in principle donate a larger number of protons than that corresponding to their
charge and can form anionic hydroxo–metal complexes.
For the case of aqueous solutions of chloroplatinic acid, Mang et al. (1993)
calculated the pH-dependent distribution of chloro–aqua and chloro–hydroxo
complexes of platinum(IV), when HCl and NaOH were used to adjust the pH. As
shown in Figure 7.3, when the pH increases the charge of the dominant species
changes from 0 to −2, while chlorine ligands are progressively replaced by H2 O
and OH− . It should be evident that the initial concentration of the precursor can
influence the equilibrium concentrations of the formed complexes.
Palladium ions in water containing chloride and/or ammonium ions form vari-
ous complex species with chloride and ammonium ligands, such as PdCl2− 4 , PdCl3
(H2 O)− , Pd(NH3 )2+4 , with electric charges ranging from −2 to +2 depending on
the solution pH (Contescu and Vass, 1987). Spielbauer et al. (1993) studied the
adsorption of palladium amino–aqua complexes on γ -Al2 O3 and SiO2 by UV spec-
troscopy, and showed that the speciation of palladium ions takes place according
to the following reactions, when pH adjustment was made by the use of ammonia:
The charge of all complexes was +2, while the number of NH3 ligands, n, increased
142 Preparation of Pellets with Nonuniform Distribution of Catalyst
7.1.2.e Coimpregnants
The presence of coimpregnants or extraneous ions in the impregnating solu-
tion can affect adsorption of the metal containing ions/complexes on the sup-
port through various mechanisms. The coimpregnants may be classified according
to the mechanism through which they affect interfacial phenomena (Heise and
Schwarz, 1987; Schwarz and Heise, 1990). The first class consists of inorganic
salts such as NaCl, NaNO3 , and CaCl2 , which affect ionic strength. The second
class includes inorganic acids and bases such as HCl, HNO3 , and NaOH, which
affect the pH of the system. These compounds can also partially dissolve the ox-
ide surface. The third class consists of compounds that compete with the metal
ion for adsorption sites. Although many chemical species can adsorb on the sur-
face, the strongest and most effective are those that contain hydroxyl, carboxyl,
and phosphoryl groups, such as acetates, citrates, and phosphates. Note that these
compounds also affect the pH and ionic strength of the solution.
Coimpregnants from the third class are commonly used to obtain nonuniform
catalyst distributions. Single component adsorption parameters of citric acid on
alumina are shown in Table 7.2. Large equilibrium adsorption constants have been
obtained (2070–19,530 l/mol), indicating strong adsorption. Surface saturation
coverage ranges from 1.43 to 3.35 µmol/m2 . Adsorption of other acids such as
acetic, lactic, oxalic and tartaric on alumina were also found to exhibit Langmuir
behavior (Engels et al., 1987; Jianguo et al., 1983).
During multicomponent adsorption, competition of the adsorbates for binding
sites takes place. Jianguo et al. (1983) showed that the amount of H2 PtCl6 adsorbed
on η-Al2 O3 was smaller in the presence of a coimpregnant. The decrease in amount
adsorbed as compared with the single-component H2 PtCl6 adsorption depended
not only on the concentration but also on the type of coimpregnant. Adsorbed
platinum in the presence of citric or tartaric acid was small, while in the presence
Area K k+ ns te
(m2 /g) (l/mol) (l/mol · s) (µmol/g) (µmol/m2 ) (h) Reference
150 215 1.43 2 Shyr and Ernst, 1980
245 820 3.35 6 Jianguo et al., 1983a
220 2,070 444 2.02 2 Engels et al., 1987
78 19,530 7.97 209 2.68 1 Papageorgiou et al., 1996
a Solid was η-Al2 O3 .
7.1 Adsorption on Powders 143
respectively. For the multicomponent adsorption rate constants, it was found that
the single-component value for hexachloroplatinic acid (2.46 l/mol · s) provided a
satisfactory representation of the transient multicomponent data. However, the
adsorption rate constant of citric acid had to be lowered to 0.2 l/mol · s in order to
achieve the best fit.
[H+ +
s ] = [H ] exp(−e'0 /kB T) (7.19)
where '0 is electrostatic potential at the surface. The exponential factor represents
the electrostatic energy required to bring a charged species from the solution bulk
to the adsorption surface. Using (7.19), equations (7.17) and (7.18) become
[S–OH][H+ ]
Ka1 = exp(−e'0 /kB T) (7.20)
[S–OH+
2]
[S–O− ][H+ ]
Ka2 = exp(−e'0 /kB T). (7.21)
[S–OH]
The material balance equation for surface sites is
NA
Ns = ([S–OH] + [S–O− ] + [S–OH+
2 ]) (7.22)
NX
where Ns is the density of surface sites (sites/m2 ). The equations above refer to
the water/oxide system. When considering the adsorption of catalytic precursors,
reactions of ions with surface sites must be included. This results in additional
equations, i.e. mass action laws related to these new surface reactions, and in
appropriate modifications of the equations above.
It is worth noting that the choice of the surface species is arbitrary and involves
the introduction of additional parameters which have to be determined by fitting
experimental data (Barrow and Bowden, 1987). A simple example is discussed
later in the context of the triple-layer model.
The surface complexation model described by the four equations (7.16), (7.20)–
(7.22) includes five unknowns ([S–OH], [S–O− ], [S–OH+ 2 ], σ0 , '0 ) and three pa-
rameters (Ka1 , Ka2 , Ns ), if the experimental pH is known. Thus, one more equation
is required. This is given by a relation between surface charge σ0 and surface po-
tential '0 , which is based on the electrostatic model employed to describe the
interface. Some models for this purpose are discussed next.
σ0 = c'0 (7.23)
146 Preparation of Pellets with Nonuniform Distribution of Catalyst
and a diffuse layer. The protonated and deprotonated surface sites and the H+ and
OH− ions are again located at the surface plane 0, and experience the potential
'0 . The IHP is separated from the surface by a region of constant capacitance c1
and contains specifically adsorbed ions at potential 'IHP . The IHP is separated
from the OHP by a region of capacitance c2 . The potential at the OHP is 'OHP and
the total diffuse layer charge is σd . The relationships between charge and potential
for the 0 plane, IHP, and OHP are
The charge in the diffuse layer is calculated by equation (7.24), where '0 is re-
placed by 'OHP . The electroneutrality condition is given by
S–O− + C+ % − +
s & S–O C (KC ) (7.33a)
S–OH+ − % + −
2 + As & S–OH2 A (KA ) (7.33b)
(where subscript s denotes the surface), the mass action equations are
[S–O− C+ ]
KC = (7.34)
[S–O− ][C+ ] exp(−e' IHP /kB T)
[S–OH+ −
2A ]
KA = . (7.35)
[S–OH+ −
2 ][A ] exp(+e'IHP /kB T)
The entire set of equations (7.20), (7.21), (7.24), (7.27)–(7.32), (7.34), (7.35) can be
solved numerically at any given pH and electrolyte concentration. The parameters
7.2 Simultaneous Diffusion and Adsorption in Pellets 149
then dry the pellet, trying not to modify it. Therefore, in this section we focus on
the impregnation process, both theoretically and experimentally, with the aim of
determining the distribution of adsorbed precursor within the pellet. We conclude
by discussing several studies which address nonuniform catalyst distributions from
both experimental and modeling viewpoints.
where Cp,i is the concentration of the ith solute in the pores, ρs is the pellet density,
and υ is the volumetric flow rate of imbibition, related to the position rf of the
advancing front by
drf
υ = −4πrf2 ε . (7.38)
dt
The derivative ∂ni /∂t represents solute uptake by the support and may be obtained
from the Langmuir adsorption equation (7.1) (for single component adsorption)
or (7.13) (for two-component competitive adsorption). In the case where adsorp-
tion is fast compared to the transport processes, the assumption of local adsorption
equilibrium can be used. Accordingly, the concentrations in the adsorbed phase
can be obtained directly from the concentrations in the pores through an adsorp-
tion model, such as for example the Langmuir isotherm equation (7.2). Therefore,
by taking the adsorption isotherm in the general form
∂ni N
∂ fi ∂Cp,k
= (7.40)
∂t k= 1
∂Cp,k ∂t
impregnation profile that would be obtained following imbibition with fast drying.
An important extension included consideration of different one-dimensional pel-
let shapes (slab, cylinder, sphere) by allowing for the effect of geometry on the
imbibition front velocity v in equation (7.41). For this purpose, a modification of
the Washburn equation, developed by Komiyama et al. (1980), was used. It was
found that the obtained nonuniform catalyst distributions degenerate to uniform
ones only when sufficient time (of the order of days) is allowed for the equilibra-
tion between the occluded liquid and the adsorbed phase. Catalyst distribution in
a pellet can also be influenced by adjusting imbibition conditions so that only par-
tial penetration of the pores is achieved. Since convective transport of solution is
due to capillary forces, critical parameters that affect the fraction of pore volume
filled with liquid are solution viscosity, surface tension, support geometry, average
pore size, and immersion time (Zhang and Schwarz, 1992).
Coimpregnation of pellets with catalyst precursor and coingredients can be
used to manipulate intrapellet distribution of the active material. The model of
Vincent and Merrill for imbibition was extended to a multicomponent system by
Kulkarni et al. (1981). A competitive Langmuir adsorption model was considered.
It was shown that when the equilibrium fractional coverage for a given concen-
tration of precursor in the liquid phase is substantially smaller in the presence of a
coimpregnant, it is possible for the precursor profile to exhibit a maximum within
the pore. This condition is equivalent to stronger adsorption of the coimpregnant
than of the catalyst precursor.
Analytical solutions for catalyst distribution at the end of the imbibition period
were obtained by Lee (1985). The model was based on competitive Langmuir
adsorption and the modified Washburn equation mentioned above for the front
velocity. Since the advantage of utilizing a coimpregnant is best realized when
adsorption is the controlling step, mass transfer resistance at the pore wall was
neglected. Egg-white distributions were obtained if both impregnants adsorbed
strongly, but the coimpregnant (species 2) adsorptivity was larger (i.e. K1 was large
and K2 > K1 ) and the rate of its adsorption was higher (k+ +
2 > k1 ) than that for the
catalyst precursor. The catalyst layer’s location and thickness can be manipulated
by adjusting the bulk concentrations of the two impregnants. The location of the
catalyst layer depends also on the pellet shape, because the latter affects the ve-
locity of the penetrating liquid front. This occurs because the liquid front moves
faster, as the pellet center is approached, in a spherical than in a cylindrical pellet;
the velocity is independent of position in a slab pellet. Under identical preparation
conditions, a slablike pellet has its shell closer to the center than a cylindrical or
spherical pellet.
In an extensive analysis of catalyst impregnation, Lee and Aris (1985) consid-
ered various one- and two-component systems with both finite and infinite adsorp-
tion rates. The latter implies instantaneous adsorption, i.e., equilibrium prevails at
any time and location throughout the wetted portion of the pellet. Under this
condition, the concentration profiles take the form of shocks moving in time. For
two-component impregnation, three regions, each with a constant composition,
are found. The more strongly adsorbed coimpregnant (species 2) is present only in
7.2 Simultaneous Diffusion and Adsorption in Pellets 153
the outer region, as illustrated in Figure 7.5. A portion of the less strongly adsorbed
precursor (species 1) passes through this region and penetrates deeper into the
pellet, where it faces no competition for the adsorption sites. Thus, the surface
concentration of adsorbed component 1 in this egg-white region is greater than
that in the eggshell region. The highly concentrated component 1 adsorbed in the
egg-white region is later washed off by the oncoming component 2, which causes
the overshoot in the concentration profile of component 1. As time progresses, the
band containing component 1 moves towards the interior of the pellet. When the
center is reached, then the egg-yolk distribution is obtained.
In the case of finite adsorption rate, the sharp concentration profiles become
blurred. Both the height and location of the catalyst layer are affected. As the rel-
ative rate of adsorption of the coimpregnant decreases, it becomes less effective in
pushing the precursor into a narrow subsurface layer. However, if the rheological
or interfacial properties of the impregnating solution are manipulated so as to
slow down the liquid imbibition rate, nonuniformities in the catalyst distribution
can be preserved.
Not only the adsorption rate but also the desorption rate affects the peak. If
the desorption rate of the precursor is decreased, its adsorbed part is washed off
slowly, smoothing the peak. Furthermore, the diffusion transport of solutes during
imbibition, superimposed on finite rates of adsorption, intensifies the blurring of
the otherwise sharp concentration profiles.
In the earlier part of this book, we have seen many instances where egg-white
catalysts provide optimal performance. It is useful to describe here the general con-
ditions which favor the formation of such distributions. The egg-white distribution
obtained by two-component imbibition, as shown in Figure 7.5, is the result of com-
petition between adsorption and convective transport. Once imbibition is over, the
catalyst deposition becomes controlled by the interaction between adsorption and
diffusion. This may either destroy or enhance the egg-whiteness of the distribution.
Lee and Aris (1985) investigated different scenarios. For example, if the two com-
ponents have same diffusivity and they are both weakly adsorbable, then the egg-
white characteristic is not pronounced, and is quickly wiped out. However, when
the precursor has larger diffusivity, egg-whiteness is modified but not destroyed. If
in addition, the adsorption of both components is sufficiently strong, egg-whiteness
is enhanced as the impregnation time is extended. As in one-component impreg-
nation, the difference between dry and wet impregnation diminishes as the im-
pregnation time is prolonged, as illustrated in Figure 7.6. Stronger adsorption of
the coimpregnant is critical in obtaining egg-white catalysts. It can effectively push
the precursor into a narrow, sharp, and concentrated band even when both com-
ponents have the same diffusivity. Larger diffusivity of the precursor makes the
overshoot of the egg-white distribution broader and higher. The adsorption ca-
pacities of the impregnants can also affect the catalyst profile. If they are small, the
catalyst distribution evolves quickly from egg-white to egg-yolk and then becomes
uniform. During this evolution, the peak is never very pronounced.
Lee and Aris (1985) have summarized the conclusions of their extensive sim-
ulations in the form of quantitative criteria for obtaining distinct egg-white or
egg-yolk catalysts, in the case of purely competitive Langmuir adsorption:
flow bring the solutes back into a region of substantial adsorption near the pore
mouth. However, the shape of the distribution can be preserved if the diffusivity
of the precursor is larger than that of the coimpregnant, because the former moves
inward faster and reaches available adsorption sites before the latter.
Hepburn et al. (1991a) studied catalyst redistribution during drying for the case
where evaporation takes place at an interface receding towards the pellet interior,
which is expected to occur in the case of fast drying (see section 7.2.2.c). The
physicochemical parameters used were representative of γ -Al2 O3 coimpregnated
with RhCl3 and HF, so as to realize an egg-white Rh distribution. The results
indicated that drying causes a buildup of RhCl3 and HF at the shrinking liquid front
where evaporation occurs. As this front moves towards the center of the support,
the increase in RhCl3 and HF concentrations in the liquid causes a competition
for unoccupied adsorption sites. RhCl3 is pushed ahead of HF because the latter
adsorbs more strongly than the former. As a result, after drying, the Rh band is
displaced towards the interior and also becomes less sharp than the Rh profile
before drying, as shown in Figure 7.8.
and Melo et al. (1980b) found that dry impregnation yields a catalyst with higher
loading and more uniformity than wet impregnation. For CuCl2 adsorption on γ -
Al2 O3 , Ott and Baiker (1983) showed that the penetration depth obtained from
dry impregnation was larger than from wet impregnation. Summers and Ausen
(1978) observed similar trends, but the difference was less significant because the
impregnation time was long.
The concentration of impregnating solution and the impregnation time deter-
mine the thickness of the shell in the case of strong adsorption. It has been observed
for various systems that extended contact with the bulk impregnating solution al-
lows the solute to diffuse into the support and enlarge the eggshell (Maatman and
Prater, 1957; Harriott, 1969; Santacesaria et al., 1977c; Baiker and Holstein, 1983;
Lee and Aris, 1985; Nicolescu et al., 1990). For Ni(NO3 )2 and Ba(NO3 )2 impreg-
nation of alumina, which is characterized by weak adsorption, wet impregnation
with short contact time, or low concentration of the impregnant, produces catalysts
with decreasing concentration profile towards the pellet center. Increasing any of
the above operating parameters smooths catalyst nonuniformities (Melo et al.,
1980b). Goula et al. (1992b) studied the impregnation of ammonium heptamolyb-
date on γ -Al2 O3 . During dry impregnation they observed that with increasing
precursor concentration, less uniform distributions were obtained. Since the op-
posite trend was expected for weak adsorption, this behavior was attributed to
a pH effect. It was confirmed that by independently increasing the pH, keeping
constant precursor concentration, more uniform distributions were obtained. In
addition, the effect of impregnation time was investigated when imbibition was
incomplete. It was found that longer impregnation times lead to smoother eggshell
profiles when imbibition was incomplete, while further increase after imbibition
was complete resulted in sharpening the eggshell distribution. This is because ad-
sorption is faster than diffusion, thus favoring precursor adsorption at the pore
mouth.
Goula et al. (1992b) did not observe any effect of the impregnating solution vol-
ume on the catalyst distribution profiles, which is reasonable for weakly adsorbing
species, since no significant precursor depletion occurs in the bulk. For the case of
strong adsorption, though, the volume of the solution would be expected to have
a similar effect as its concentration, since it would affect the amount of precursor
available for adsorption. Even for long impregnation times, since desorption is
slow, the precursor is distributed nonuniformly if the amount in solution is less
than that required to saturate, whereas if there is an adequate amount of solute,
a uniform distribution can be obtained (Maatman, 1959).
The impregnating solution’s pH and ionic strength have an indirect but signifi-
cant effect on adsorption characteristics of the precursor. By altering the pH in the
direction which favors adsorption of the impregnating species, sharper eggshell
distributions can be obtained (Maitra et al., 1986; Heise and Schwarz, 1985; Goula
et al., 1992b). For the system H2 PtCl6 /γ -Al2 O3 , when the ionic strength of the im-
pregnating solution was increased by addition of simple inorganic salts, Heise
and Schwarz (1986) found that the penetration depth of the eggshell distribution
was not affected, but the adsorbed precursor concentration decreased and the
7.2 Simultaneous Diffusion and Adsorption in Pellets 161
the evaporation front is continuous and recedes towards the pellet center as dry-
ing progresses, while in slow drying the evaporation front is initially located at
the pellet surface, later portions of it move inward, and finally it breaks down
so that isolated domains of liquid phase are formed (Neimark et al., 1976, 1981;
Fenelonov et al., 1979). Evaporation at the pellet surface and at a moving front
receding towards the pellet interior, depending on the drying conditions, have
been experimentally observed (Hollewand and Gladden, 1992; Gladden, 1995).
In addition, it has been shown that intrapellet moisture distribution can be quite
different for pellets with different pore size distribution even when the overall
drying rate is the same (Hollewand and Gladden, 1994).
The redistribution of solute is affected not only by liquid capillary flow but also
by diffusive transport of the solute. The nonvolatile solute concentration increases
in regions where the liquid vaporizes, thus creating concentration gradients, which
cause diffusion of the solute (Lee and Aris, 1985). If it is fast enough as compared
to vapor removal, diffusional relaxation smooths out concentration gradients. In
the opposite case, high solute concentration exists at the evaporation front, and
if it exceeds the saturation concentration, crystallites are formed. The interac-
tion between mass transport and crystallite nucleation/growth can affect solute
concentration gradients and catalyst dispersion, as discussed in section 2.4.1.
From the above discussion it follows that in general the solute distribution is
determined by the relative magnitudes of evaporation rate, capillary flow, and so-
lute diffusion. Since capillary flow is much faster than diffusive transport, for the
purposes of precursor redistribution it is more appropriate to define as fast drying
the regime where liquid evaporation is much faster than capillary flow (Jv Jc ),
as slow drying the regime where liquid evaporation is much slower than diffusive
transport (Jd Jv ), and as intermediate drying the regime where liquid evapo-
ration is much slower than capillary flow but much faster than diffusive transport
(Jc Jv Jd ). With respect to the final precursor concentration profile in the
pellet, fast drying allows one to preserve the same profile present in the pore-filling
liquid at the end of the impregnation step, i.e., no redistribution of the precursor
occurs. In all other cases the precursor concentration profile in the pore liquid
changes during drying and consequently alters the deposited precursor profile
in the pellet. The final profile is the result of complex interactions among the
transport processes mentioned above, and is also affected by adsorption and pre-
cipitation in addition to the local and global pore size distribution and connectivity.
Other processes such as solvent adsorption, film flow (i.e. two-phase capil-
lary flow) and recondensation through the gas phase may affect liquid transfer
(Fenelonov, 1975; Neimark et al., 1981). Because of the large number of param-
eters and their interactions, it is difficult to draw conclusions of general valid-
ity about the effects of drying on catalyst distribution. Accordingly, conclusions
reported in the literature, particularly in experimental work, are limited to the
specific systems and conditions investigated. In the following we report some of
these, to illustrate the effects of drying on catalyst distribution, although, owing to
lack of quantitative data about diffusion and capillary flow, they do not provide a
comprehensive picture.
7.2 Simultaneous Diffusion and Adsorption in Pellets 167
The effect of drying has also been studied in cases where at the end of impreg-
nation an egg-white catalyst distribution is obtained. As expected, the active layer
is preserved by fast drying, since the adsorbed precursor does not have time to
redistribute (Hepburn et al., 1989b). Milder drying conditions can cause blurring
of the catalyst layer, which becomes diffuse and wider and moves inwards (Lee
and Aris, 1985). In addition, Hepburn et al. (1989b) and Becker and Nuttall (1979)
noticed that in such cases, catalyst became concentrated also towards the external
surface, giving rise to bimodal distributions, as shown in Figure 7.12. By drying
slowly at 20◦ C, all peaks were completely washed out, giving a uniform catalyst
(Hepburn et al., 1989b). In light of the discussion earlier in this section, it appears
that drying at 20◦ C corresponds to slow, at 300 and 400◦ C to intermediate, and at
500◦ C to fast drying. Finally, two other alternatives for minimizing catalyst redis-
tribution during drying are worth mentioning: immobilization of precursor before
7.2 Simultaneous Diffusion and Adsorption in Pellets 169
drying through liquid-phase reduction (Harriott, 1969; Becker and Nuttall, 1979;
Kunimori et al., 1982), and freeze drying with liquid nitrogen (Chu et al., 1989).
Only the imbibition period was considered. The convective flow rate was evaluated
using the Washburn equation, and diffusive transport was neglected. By fitting
experimental catalyst profiles, they concluded that adsorption was controlled by
kinetics and not by mass transfer. The model was consistent with observed nickel
profiles, for both one- and two-component systems.
Heise and Schwarz (1988) prepared platinum catalysts by dry impregnation of
γ -Al2 O3 pellets with hexachloroplatinic acid. Adsorption equilibrium data were
obtained independently and shown to follow a Langmuir isotherm. In this case
also, the impregnation model (7.41), (7.42) was used, assuming adsorption kinetics
to control the precursor flux from the impregnating solution to the pore wall. The
adsorption kinetic parameters were evaluated by fitting platinum uptake during
the impregnation experiments. The calculated platinum concentration profiles
in the pellets were shown to be in qualitative agreement with the experimental
eggshell profiles.
Platinum profiles obtained after coimpregnation of γ -Al2 O3 with hexachloro-
platinic and phosphoric acids were also simulated (Schwarz and Heise, 1990). For
this, a multicomponent single-pore model for dry impregnation as discussed in
section 7.2.1.a was used (Kulkarni et al., 1981). Phosphoric acid adsorption was
assumed to follow Langmuir kinetics with an adsorption constant three times
greater than that for the platinate ion. Qualitative agreement was obtained only
after effects of the coimpregnant on solution-interface chemistry were incorpo-
rated into the model, viz., the influence of decreasing pH was simulated by de-
creasing the adsorption capacity of the support.
Hepburn et al. (1991a) impregnated dry γ -Al2 O3 honeycombs with rhodium
trichloride and hydrofluoric acid, and obtained egg-white Rh catalyst distributions.
For both impregnants, single-component equilibrium adsorption data were mea-
sured and fitted with Langmuir adsorption isotherms. Multicomponent adsorption
was described by the competitive Langmuir model [see equation (7.13)]. The cata-
lyst preparation was simulated by considering separately the two steps: imbibition
and drying. For the first, the single-pore model (7.41), (7.42) was used, assuming ad-
sorption kinetic control; the second was simulated through a diffusion–adsorption
model, similar to equation (7.44), with appropriate BCs to take account of sol-
vent evaporation. Two adjustable parameters were used, viz. the adsorption rate
constant of RhCl3 and the effective diffusivity of HF. The two models were used
sequentially to reproduce the Rh profile measured experimentally. Good quanti-
tative agreement was obtained by tuning the two adjustable parameters, as shown
in Figure 7.13. The prediction of the peak height was particularly sensitive to
the chosen value of the RhCl3 adsorption rate constant, while the peak location
was primarily controlled by the surface saturation constant determined from the
adsorption equilibrium data.
and irreversible, so that the time scale for adsorption is much smaller than that for
diffusion, the model reduces to a shell progressive model. This has been used to
simulate experimental results of CuCl2 (Baiker and Holstein, 1983) and H2 PtCl6
impregnation of γ -Al2 O3 (Santacesaria et al., 1977c). In both cases, effective dif-
fusivities were estimated by fitting the catalyst penetration depth as a function of
time. Baiker and coworkers (Baiker and Holstein, 1983; Baiker et al., 1986) took
into account both adsorbed copper and copper precipitated from the pore solution
during drying, which ultimately gave rise to two types of copper species with dif-
ferent properties as identified by TPR. The agreement between experimental and
calculated catalyst profiles was good, even though the model predicted a slower
approach of the penetration front to the center of the pellet.
Hanika and coworkers (Hanika et al., 1982, 1983; Machek et al., 1983a) pre-
pared Pt catalysts on activated carbon and alumina, utilizing aqueous and or-
ganic solutions of hexachloroplatinic acid, as well as phosphene and amine com-
plexes of platinum. The one-component wet-impregnation model (7.44), (7.45)
was used, assuming local adsorption equilibrium according to Langmuir or Fre-
undlich isotherms. The effective diffusivities were calculated by fitting the time-
dependent Pt uptake from the impregnating solution. The computed catalyst dis-
tributions were in qualitative agreement with the experimental data.
A two-component (RhCl3 –HF) wet-impregnation system was considered by
Hegedus et al. (1979b) using the adsorption–diffusion model (7.44). The multi-
component adsorption was described by the competitive Langmuir model (7.13),
whose parameters were estimated from single-component data of adsorption ki-
netics on powder support. In the Langmuir model, the various species were al-
lowed to have different saturation concentrations. Diffusivities were determined
by comparing the measurements of species uptake on pellets, using results of the
diffusion–adsorption model. The model predicted Rh peak locations inside the
γ -Al2 O3 pellets satisfactorily.
7.2 Simultaneous Diffusion and Adsorption in Pellets 173
A similar approach was followed by Scelza et al. (1986) to simulate wet coim-
pregnation of γ -Al2 O3 pellets with H2 PtCl6 and HCl. The same competitive
Langmuir model was used to describe adsorption kinetics, but now assuming
the same saturation concentration for all components. Moreover, the diffusion
coefficients and the adsorption rate constants were estimated simultaneously by
fitting experimental data on species uptake from the impregnating solution as
a function of time. Unfortunately, different values of the adjustable parameters
were obtained at each acid concentration, although the scatter was significant
only for desorption rate constants. Modeling for various acid concentrations was
carried out utilizing the corresponding set of diffusivities and adsorption param-
eters, and good agreement between theoretical and experimental profiles was
obtained. It was observed that with increasing HCl concentration the profile de-
velops from eggshell to uniform. The same approach was also used to simulate
bimetallic wet coimpregnation. In particular, experimental data on H2 PtCl6 and
HReO4 coimpregnation in the presence of HCl were considered (De Miguel et al.,
1984; Ardiles et al., 1986). The predicted and experimental profiles of the various
species after 12-h impregnation were in good agreement except the Re profiles,
for which the predicted trend to form an egg-white distribution was not confirmed
experimentally. Finally, it was found that the model was also able to predict Pt,
Re, and Cl profiles obtained by dry impregnation, indicating that the imbibition
period had no important influence due to the long impregnation time used in the
experiments.
Melo et al. (1980b) prepared Ni and Ba catalysts on γ -Al2 O3 by wet and dry
impregnation. They performed single- and two-component experiments, where
Ni(NO3 )2 and Ba(NO3 )2 were impregnated simultaneously or successively. In
the latter case, Ni was impregnated first, and then, after calcination, Ba was im-
pregnated as the second component. Wet impregnation was simulated with the
adsorption–diffusion model (7.44), introducing external resistance to mass trans-
fer and assuming local adsorption equilibrium inside the pellet. This was described
through the multicomponent Langmuir model, whose parameters were estimated
from independent equilibrium data (Melo et al., 1980a). The results of the wet-
impregnation model were used as initial conditions for the drying model, to take
proper account of the effect of drying on catalyst distribution. For dry impregna-
tion a simplified model was used to describe the first imbibition period, and the
wet-impregnation model was used again for the subsequent part of the process.
The simplified model was based on the assumption that the penetration rate is
much higher than the adsorption rate, so that adsorption could be neglected dur-
ing imbibition (Melo et al., 1980b). From the first set of experiments involving
one-species impregnation, it was found that diffusion and external mass trans-
fer coefficients could not be evaluated independently but had to be estimated by
fitting the species uptake data as a function of time. On the other hand, these
parameter values allowed them to predict with reasonable accuracy the catalyst
profiles obtained in two-species impregnation experiments. It is worth noting that
to improve agreement with the experimental data, an adjustable parameter was
introduced. This was the fraction of adsorption sites postulated to be accessible
174 Preparation of Pellets with Nonuniform Distribution of Catalyst
to nickel, while the remaining ones were available for both metals. On the whole,
experimental profiles for one-component impregnation exhibited a weak eggshell
shape, while the calculated ones were sharper. Two-component impregnations
yielded more uniform distributions.
In the above studies, effective diffusivities were estimated by comparing ex-
perimental solute uptake measurements in pellets with results of a diffusion–
adsorption model. In some cases, not only diffusivities but also other parameters
(e.g. adsorption constants, external mass transfer coefficients) were adjusted to
fit experimental data. Price and Varma (1988) developed an experimental proce-
dure for the independent evaluation of diffusivity, based on the idea of decou-
pling adsorption and diffusion. This was done by running diffusion experiments
at sufficiently high solute concentrations. In Langmuir adsorption, surface con-
centration of adsorbate does not change appreciably at saturation conditions, i.e.
for fluid concentration values beyond a sufficiently high value. Thus, on keeping
both the pore and the bulk fluid in this concentration range, adsorption and des-
orption do not occur. Hence, the diffusion–adsorption model equations (7.44),
(7.45) are simplified, in that, ∂n/∂t = 0. In order to ensure saturation conditions,
pellets were placed in a high-concentration solution for several days, prior to
their use in diffusivity experiments. They were then moved to a solution of dif-
ferent (but still sufficiently high) concentration. Effective diffusivities were deter-
mined by solving equations (7.44), (7.45) and calculating the value of De which
gave the best fit to the time-dependent solute concentration change in the bulk
solution.
Papageorgiou et al. (1996) prepared Pt/γ -Al2 O3 pellets with a sharp internal
step distribution of catalyst using coimpregnation and sequential impregnation of
hexachloroplatinic and citric acids. In the sequential impregnation, the catalyst
precursor was impregnated first, and then the competitive adsorbate. The latter
adsorbed more strongly and pushed the former towards the center of the pellet,
resulting in egg-white catalysts. Impregnation in the reverse order also produced
similar egg-white catalysts, because the catalyst precursor moved past the pellet
portion where citric acid had adsorbed (Papageorgiou, 1984). Similar behavior was
observed for γ -Al2 O3 pellets impregnated in a sequential fashion with chromium
nitrate (strongly adsorbed) and copper nitrate (less strongly adsorbed) (Chen
et al., 1980).
The cross sections of pellets prepared by sequential and coimpregnation tech-
niques are shown in Figure 7.14. The dark rings in Figure 7.14.a–i are the platinum
catalyst, while the lighter portion is the γ -Al2 O3 support. For comparison, eggshell,
egg-yolk, and uniform pellets are also shown in Figure 7.14.j–l. For both types of
impregnations, the platinum band was located deeper inside the pellet for higher
initial citric acid concentration (compare Figure 7.14.b–c, e–f, h–i). In addition,
longer impregnation time forced the platinum to move closer to the pellet cen-
ter, since the citric acid could diffuse further inside the pellet and displace the
adsorbed platinum (compare Figure 7.14.a–b, d–e, g–h).
Increasing the initial hexachloroplatinic acid concentration did not have any
noticeable effect on catalyst location; however, it resulted in catalysts with higher
7.2 Simultaneous Diffusion and Adsorption in Pellets 175
loading and somewhat larger bandwidths (compare Figure 7.14.a–d, b–e, c–f).
The bandwidth was also influenced by the impregnation time, owing primarily to
geometric effects. Since the diameter of the ring where the platinum was located
decreased with time, its width increased to accommodate the deposited metal
(compare Figure 7.14.a–b, d–e, g–h). In general, for the same impregnation time
and adsorbate concentrations, similar penetration depths were achieved by the
176 Preparation of Pellets with Nonuniform Distribution of Catalyst
two types of impregnations (compare Figure 7.14.a–g, b–h), but the bandwidth
was smaller (as small as 50 µm) when using sequential impregnation.
The eggshell distribution (Figure 7.14.j) was prepared by impregnating the pel-
let in a solution of hexachloroplatinic acid alone. The egg-yolk distribution (Figure
7.14.k) could be obtained by both sequential impregnation and coimpregnation,
provided that the impregnation periods were sufficiently long. The uniform distri-
bution (Figure 7.14.l) was prepared using the coimpregnation method. Note that
the difference in the preparation of pellets depicted in Figure 7.14.k and l is only in
the impregnation duration. Thus, for short coimpregnation time an egg-white de-
position is obtained, while for longer impregnation times egg-yolk and ultimately
uniform distributions are attained.
Modeling of coimpregnation and sequential impregnation of pellets was car-
ried out using equations (7.44), (7.45). Figures 7.15 and 7.16 show the predicted
dimensionless platinum surface concentration profiles, along with the experimen-
tally measured platinum band positions denoted by the shadowed regions for the
sequential-impregnation and coimpregnation studies, respectively. The numbers
on the bands give the impregnation time (for sequential impregnation it is the
time in citric acid solution). The penetration depth of platinum predicted by the
7.2 Simultaneous Diffusion and Adsorption in Pellets 177
even after long impregnation times, due to the slow migration of protons rela-
tive to the metal ions. By controlling the initial pH profile, eggshell, egg-white,
egg-yolk, and uniform catalysts can be obtained. For example, for producing egg-
white catalysts, an acid pretreatment is required to yield a region of lower pH at
the pellet surface than in the interior, followed by nickel impregnation. Nickel
ions diffuse relatively quickly past the lower pH shell and adsorb as they reach the
higher-pH internal region, creating an egg-white profile. For longer impregnation
times, nickel diffusion continues and the egg-white region enlarges to produce an
egg-yolk catalyst. Agreement of the model calculations with experimental nickel
distributions was good in most cases, as shown in Figure 7.18.
APPENDIX A
181
182 Application of the Maximum Principle for Optimization
These equations represent both the pellet and the membrane problems through
the following assignments:
Pellet problem.
Objective Effectiveness factor Selectivity Yield
function η S Y
x1 x x x
x2 γ γ γ
x3 h h h
x4 0 xN 0
x5 0 xM 0
f1 γ s −n γ s −n γ s −n
µ(s) µ(s) µ(s)
f2 w(x)s n w(x)s n w(x)s n
1 + Bµ(s) 1 + Bµ(s) 1 + Bµ(s)
f3 µ(s) s n µ(s) s n µ(s) s n
µ(s)
f4 0 w N (x)s n 0
1 + Bµ(s)
µ(s)
f5 0 w M (x)s n 0
1 + Bµ(s)
f xf xf xf
I ηM SNM YNM
DM xN (1)
G 0 0
DN xM (1)
µ(s) µ(s)
R w M (x)s n 0 w N (x)s n
1 + Bµ(s) 1 + Bµ(s)
Membrane problem.
Objective Conversion Selectivity Purity
function X S P
x1 x x x
x2 γ γ γ
x3 h h h
x4 0 0 0
x5 0 0 0
f1 γ s −n γ s −n γ s −n
µ(s) µ(s) µ(s)
f2 w(x)s n w(x)s n w(x)s n
1 + Bµ(s) 1 + Bµ(s) 1 + Bµ(s)
f3 µ(s) s n µ(s) s n µ(s) s n
f4 0 0 0
f5 0 0 0
f xf = p3 [γ(1)] xf = p3 [γ(1)] xf = p3 [γ(1)]
I XN SDN PDN
G (B.27) (B.30) (B.32)
R 0 0 0
Application of the Maximum Principle for Optimization 183
The optimization problem can be solved using the variational method described
in more general terms in Ray (1981). Here we show its application to the cases of
interest. Let us assume that µ̄(s) is the optimal control function and express any
other control function infinitesimally close to it as a perturbation about µ̄(s):
The state x(s) resulting from µ(s) can also be represented as a perturbation about
the state x̄(s) caused by the control function µ̄(s):
Next we linearize equations (A.1)–(A.5) about the optimal function µ̄, to yield
d(δx1 ) ∂fT
= 1 δx2 (A.16)
ds ∂x2
d(δx2 ) ∂fT ∂f2
= 2 δx1 + δµ (A.17)
ds ∂x1 ∂µ
d(δx3 ) ∂ f3
= δµ (A.18)
ds ∂µ
d(δx4 ) ∂ f4 ∂ f4
= δx1 + δµ (A.19)
ds ∂x1 ∂µ
d(δx5 ) ∂ f5 ∂ f5
= δx1 + δµ. (A.20)
ds ∂x1 ∂µ
Let us now adjoin to the objective functional the linearized constraints [equations
(A.16)–(A.20)] multiplied by the adjoint variables (Lagrange multipliers) λ i (s):
1
d(δx1 ) ∂ f1T
λT1 (s) − δx2 ds = 0 (A.22)
0 ds ∂x2
1
d(δx2 ) ∂f2T ∂f2
λT2 (s) − δx1 − δµ ds = 0 (A.23)
0 ds ∂x1 ∂µ
1
d(δx3 ) ∂ f3
λ3 (s) − δµ ds = 0 (A.24)
0 ds ∂µ
1
d(δx4 ) ∂ f4 ∂ f4
λ4 (s) − δx1 − δµ ds = 0 (A.25)
0 ds ∂x1 ∂µ
1
d(δx5 ) ∂ f5 ∂ f5
λ5 (s) − δx1 − δµ ds = 0. (A.26)
0 ds ∂x1 ∂µ
184 Application of the Maximum Principle for Optimization
∂G ∂G ∂G
δI = δx2 |1 + δx4 |1 + δx5 |1
∂x2 ∂ x4 ∂ x5
1 1
∂R ∂f T
∂ f4 ∂ f5 ∂fT
+ + λT2 2 + λ4 + λ5 δx1 ds + λT1 1 δx2 ds
0 ∂x1 ∂x1 ∂x1 ∂x1 0 ∂x2
1
∂R ∂f2 ∂ f3 ∂ f4 ∂ f5
+ + λT2 + λ3 + λ4 + λ5 δµ ds
0 ∂µ ∂µ ∂µ ∂µ ∂µ
1
d(δx1 ) d(δx2 ) d(δx3 ) d(δx4 ) d(δx5 )
− λT1 + λT2 + λ3 + λ4 + λ5 ds.
0 ds ds ds ds ds
(A.27)
Substituting the above equations (A.28a)–(A.28b) and using the initial conditions
(A.9)–(A.12), the first and last terms of the rhs of equation (A.27) become
∂G 1
d(δx1 ) d(δx2 ) d(δx3 ) d(δx4 ) d(δx5 )
δx2 |1 − λT1 + λT2 + λ3 + λ4 + λ5 ds
∂x2 0 ds ds ds ds ds
∂G
= δx2 |1 − λT1 δx1 |1 + λT1 δx1 |0 − λT2 δx2 |1 + λT2 δx2 |0 − λ4 δx4 |1 − λ5 δx5 |1
∂x2
1
dλT1 dλT2 dλ3 dλ4 dλ5
+ δx1 + δx2 + δx3 + δx4 + δx5 ds. (A.29)
0 ds ds ds ds ds
δx2 |0 = 0 (A.31)
and
δx2 |1 = A J1 δx2 |1 − δx1 |1 (A.32)
and then
−1
δx2 |1 = − I − A J1 A δx1 |1 . (A.33)
We want this expression to be zero for all δx1 |0 and δx1 |1 ; thus we set
λ1 = 0 at s=0 (A.37)
T
T ∂G
λ2 = A−1 λ1 − JT1 λ1 + at s = 1. (A.38)
∂x2
Note that since for the pellet problem neither f nor G is a function of x2 , the BC
(A.38) becomes
T
λ2 = A−1 λ1 at s = 1
186 Application of the Maximum Principle for Optimization
which is equivalent to
1
∂ H dλ3 1
∂ H dλ4
+ + δx3 ds + + δx4 ds
0 ∂ x3 ds 0 ∂ x4 ds
1
∂ H dλ5 1
∂H
+ + δx5 ds + δµ ds
0 ∂ x5 ds 0 ∂µ
∂G ∂G
+ − λ4 δx4 |1 + − λ5 δx5 |1 (A.40)
∂ x4 ∂ x5
where
is called the Hamiltonian. In order to express the direct influence of δµ alone, let
us choose the arbitrary functions λ such that they satisfy
dλT1 ∂H
=− (A.42)
ds ∂x1
dλT2 ∂H
=− (A.43)
ds ∂x2
dλ3 ∂H
=− (A.44)
ds ∂ x3
dλ4 ∂H
=− (A.45)
ds ∂ x4
dλ5 ∂H
=− . (A.46)
ds ∂ x5
This allows the influence of the system equations (A.1)–(A.5) to be transmitted by
Application of the Maximum Principle for Optimization 187
λ and to be felt in ∂H/∂µ. For the terms outside the integral in (A.40) we choose
∂G
λ4 = at s=1 (A.47)
∂ x4
∂G
λ5 = at s = 1. (A.48)
∂ x5
Thus, on using equations (A.42)–(A.48), equation (A.40) becomes
1
∂H
δI = δµ ds. (A.49)
0 ∂µ
In this appendix, equations describing the mass and energy balances in pellets and
in both feed and permeate-side fluid streams of an inert membrane reactor, shown
in Figure 5.1, are presented (cf. Yeung et al., 1994). The problem of determining
the optimal catalyst distribution in pellets is then formulated. Optimization of the
catalyst loading profile is carried out, as in section 2.5.1, using as optimization
variable the dimensionless loading of the active catalyst
q(s)
µ(s) = (B.1)
q¯
which, by definition, satisfies the condition
1
1
µ(s) s n ds = . (B.2)
0 n + 1
On physical grounds, it is assumed that µ(s) is positive and bounded, i.e.
0 ≤ µ(s) < α (B.3)
where α represents the maximum allowable value of the local catalyst loading on
the support.
The dependence of the reaction rate on catalyst loading is represented by equa-
tion (2.129):
µ(s)
r = r0 . (B.4)
1 + Bµ(s)
(B.5)
188
Optimal Catalyst Distribution in Pellets for an Inert Membrane Reactor 189
J
1 d n dθ µ(s)
L[θ] = n s =− βj φ 2j fj (u, θ ) (B.6)
s ds ds 1 + Bµ(s) j=1
with BCs
dui
s = 0: =0 (i = 1, . . . , I) (B.7a)
ds
dui
s = 1: = Bim,i (gi − ui ) (i = 1, . . . , I) (B.7b)
ds
dθ
s = 0: =0 (B.7c)
ds
dθ
s = 1: = Bih (τ − θ ) (B.7d)
ds
where the following dimensionless parameters are used:
ui = Ci /C f,1
0
, s = x/R, Di = De,1 /De,i
gi = Cf,i /C f,1
0
, τ = Tf /Tf0
φ 2j = r j Cf,1
0
, Cf,1
0
, . . . , Cf,1
0
, Tf 0 R 2 /De,1 Cf,1
0
0 (B.8)
f j (u, θ) = rj (C1 , C2 , . . . , CI , T)/r j Cf,1 , Cf,1
0
, . . . , Cf,1
0
, Tf 0
θ = T/Tf 0 , β j = (−Hj )De,1 Cf,1 0
/ λe Tf 0
Bim,i = kg,i R/De,i , Bih = hR/λe .
The mass and energy balances can be rewritten in matrix form as follows:
µ(s)
L[x] = w(x) (B.9)
1 + Bµ(s)
with BCs
dx
s = 0: =0 (B.10a)
ds
dx
s = 1: = A(xf − x) (B.10b)
ds
where the vectors x, xf , and w, and the matrix A, are defined by
J
D1 ν1 j φ j f j (x)
2
j=1
J
u1 g1
D2 ν2 j φ 2j f j (x)
u2
j=1 g2
.. . .
x = . , w (x) = .. ,
xf = .. (B.11)
uI J gI
DI ν I j φ j f j (x)
2
θ τ
j=1
J
− β φ 2 f (x)
j j j
j=1
190 Optimal Catalyst Distribution in Pellets for an Inert Membrane Reactor
Bim,1 0
0 Bim,2 0
A=
..
.
..
.
..
. .
(B.12)
0
0 Bim,I
0 Bih
The steady-state mass balances for the cylindrical inert membrane, accounting
for external transport resistances, are given by
1 d dCM,i
DeM,i xM =0 (i = 1, . . . , I) (B.13)
xM dxM dxM
with BCs
dCM,i
xM = xF : −DeM,i = kg,F (CF,i − CM,i ) (B.14a)
dxM
dCM,i
xM = xP : −DeM,i = kg,P (CM,i − CP,i ). (B.14b)
dxM
The mass flux of component i, at the inner or outer membrane surface, can be
obtained on solving the above equations:
dCM,i dCM,i
−2π xP LDeM,i = −2π xF LDeM,i
dxM xM =xP dxM xM =xF
ζF,i PF ζP,i PP
2π LDeM,i Rg TF
− Rg TP
= . (B.15)
ln xP
xF
+ DeM,i kg,F1 xF + kg,P1 xP
2π LλeM (TF − TP )
= . (B.16)
ln xP
xF
+ λeM hF1xF + hP1xP
The mass balances for the feed side of the membrane reactor are given by
Vr (1 − ε)(n + 1)De,i CF,1
0
dui (1)
QF0 ζF,i
0
− QF ζF,i −
R2 ds
ζF,i PF ζP,i PP
2π LDeM,i Rg TF
− Rg TP
− 1 =0 (B.17)
ln xP
xF
+ DeM,i kg,F xF + kg,P1 xP
Additionally, we have the two congruence equations arising from the definition
of the mole fractions in the feed and permeate streams,
CF,i CP,i
ζF,i = I , ζP,i = I (B.19)
i=1 CF,i i=1 CP,i
given by
I
I
ζF,i = 1 and ζP,i = 1. (B.20)
i=1 i=1
The energy balance for the feed side of the membrane reactor is
2π LλeM (TF − TP )
ρcp QP0 T0
TP0 P
− (ρcp QP )TP TP − = 0. (B.22)
ln + λeM hF1xF + hP1xP
xP
xF
ζ F = p1 [γ(1)] (B.23)
ζ P = p2 [γ(1)] (B.24)
xf = p3 [γ(1)] (B.25)
QF0 ζF,N
0
+ QP0 ζP,N
0
− QF ζF,N − QP ζP,N
XN = (B.26)
QF0 ζF,N
0
+ QP0 ζP,N
0
192 Optimal Catalyst Distribution in Pellets for an Inert Membrane Reactor
KN γ N (1)
XN = (B.27)
QF ζF,N + QP0 ζP,N
0 0 0
where
Vr (1 − ε)(n + 1)De,N CF,1
0
KN = − . (B.28)
R2
Selectivity of a product D relative to a reactant N:
QF ζF,D + QP ζP,D
SDN = (B.29)
QF0 ζF,N
0
+ QP0 ζP,N
0
− QF ζF,N − QP ζP,N
QF0 ζF,D
0
+ QP0 ζP,D
0
− KDγD(1)
SDN = . (B.30)
KN γN (1)
ζF,D
PDN = (B.31)
ζF,N
p1D [γ(1)]
PDN = = D[γ(1)]. (B.32)
p1N [γ(1)]
dx
= γ s −n (B.33)
ds
dγ µ(s)
= w(x) s n (B.34)
ds 1 + Bµ(s)
dh
= µ(s) s n (B.35)
ds
with BCs
s = 0: γ=0 (B.36a)
s = 1: γ = A(xf − x) (B.36b)
s = 0: h=0 (B.36c)
1
s = 1: h= . (B.36d)
n+1
Optimal Catalyst Distribution in Pellets for an Inert Membrane Reactor 193
We can now introduce the Hamiltonian, whose expression for all performance
indexes is
µ(s) T
H= λ2 w(x) s n + λT1 γs −n + λ3 µ(s) s n (B.37)
1 + Bµ(s)
where λ1 and λ2 (column vectors, with I + 1 elements) and λ3 are the Lagrange
multipliers defined by the following differential equations:
dλ1 µ(s)
=− s n [JT λ2 ] (B.38)
ds 1 + Bµ(s)
dλ2
= −λ1 s −n (B.39)
ds
dλ3
=0 (B.40)
ds
where J is the Jacobian matrix of the vector w with respect to the independent
variable x. The relevant BCs for the Lagrange multipliers are
s = 0: λ1 = 0 (B.41a)
−1 T
s = 1: λ2 = (A ) λ1 − JT1 λ1 +v (B.41b)
vi = 0 for i = N (B.42a)
KN
vN = . (B.42b)
QF0 ζF,N
0
+ QP0 ζP,N
0
vi = 0 for i = N, D (B.43a)
QF0 ζF,D
0
+ QP0 ζP,D
0
− KDγD(1)
vN = − (B.43b)
KN γ N2 (1)
KD
vD = − . (B.43c)
KN γN (1)
For product purity maximization
v = dγ (B.44)
A surface area of active element per unit catalyst pellet weight (m2 per gram of
catalyst pellet)
A matrix defined by equation (2.102)
Ai component involved in a reaction
a dimensionless catalyst activity
ap pore surface area per unit volume
B bq̄
Bi∗ dimensionless parameter defined by equation (2.32)
Bih Biot number for heat transfer, hR/λe
Bim Biot number for mass transfer, kg R/De
b constant defined by equation (2.122)
b1 , b2 constants defined by equation (3.27)
C concentration
c concentration vector
c capacitance
cp fluid heat capacity
D catalyst dispersion
De effective diffusivity
Di dimensionless effective diffusivity
D0 catalyst dispersion as loading q → 0
Da Damköhler number
d number of zones in a reactor
E activation energy
e reactant consumption rate defined by equation (3.11)
e elementary charge
F Faraday constant or flux
f, f j dimensionless reaction rate
f¯ dimensionless production rate of desired product, integrated over the
catalyst pellet
G∗ function defined by equation (2.80)
G performance index for deactivating catalyst defined by equation (4.12)
195
196 Notation
r, r j , r reaction rate
r̄ volume-averaged reaction rate defined by equation (3.3)
rf position of imbibition front
r0 reaction rate in a catalyst pellet where the active surface area per unit
catalyst weight is equal to pq̄
S selectivity
St Stanton number
s dimensionless radial coordinate
s̄ dimensionless location of catalyst in pellet
T temperature
t time
U1 , U2 objective functions defined by equations (2.108), (2.109)
u dimensionless solid-phase concentration
ū dimensionless concentration at the catalyst location
u dimensionless solid-phase concentration vector
V volume
Vp pellet volume
v velocity or bond valence
W catalyst weight or mass flux
w dimensionless feed distribution along a reactor
w vector defined by equation (2.101)
w Mx vector defined by equation (2.143)
X reactant conversion
XT total conversion for membrane reactor defined by equation (5.1)
x radial coordinate
x vector defined by equation (2.101)
xf vector defined by equation (2.101)
Y yield
y dimensionless axial coordinate
ŷ reactor portion with subsurface optimal location or reactor portion
with distributed feed
Z ion valence
z axial coordinate
[ ] concentration
Greek Letters
α upper bound of the dimensionless active-element concentration
defined by equation (2.128)
α1 weighting coefficient proportional to value of products
α2 weighting coefficient proportional to cost of raw materials
α3 weighting coefficient proportional to cost of catalyst
αg constant [see equation (5.14)]
β dimensionless heat of reaction or fraction of sulfonic groups
β̄ external Prater number
198 Notation
Subscripts
0 0 plane; initial
b bulk liquid
c coolant
d diffuse layer
e equilibrium
ext external
F feed section of a membrane reactor
f fluid phase
IHP inner Helmholz plane
i component i
ip for product inhibition
is for substrate inhibition
j reaction j
M membrane
m, max maximum
OHP outer Helmholz plane
oper operating
opt optimal
P permeate
p poison or pore
r reactor
s saturation or surface
un uniform
Superscripts
0 reactor inlet
* value corresponding to the optimal catalyst distribution
References
Abello, M. C., Velasco, A. P., Gorriz, O. F., and Rivarola, J. B. 1995. Temperature-
programmed desorption study of the acidic properties of γ -alumina, Appl. Catal. A.
129: 93–100.
Aben, P. C. 1968. Palladium areas in supported catalysts. Determination of palladium sur-
face areas in supported catalysts by means of hydrogen chemisorption, J. Catal. 10:
224–229.
Adcock, D. S., and McDowall, I. C. 1957. The mechanism of filter pressing and slip casting,
J. Amer. Ceram. Soc. 40: 355–362.
Agashe, K. B., and Regalbuto, J. R. 1997. A revised physical theory for adsorption of metal
complexes at oxide surfaces, J. Coll. Interf. Sci. 185: 174–189.
Ahn, J. H., Ihm, S. K., and Park, K. S. 1988. The effects of the local concentration and
distribution of sulfonic acid groups on 1-butene isomerization catalyzed by macroporous
ion-exchange resin catalyst, J. Catal. 113: 434–443.
Akratopulu, K. C., Vordonis, L., and Lycourghiotis, A. 1986. Effect of temperature on
the point of zero charge and surface dissociation constants in aqueous suspensions of
γ -Al2 O3 , J. Chem. Soc., Faraday, Trans. I 82: 3697–3708.
Akratopulu, K. C., Vordonis, L., and Lycourghiotis, A. 1988. Development of carriers with
controlled concentration of charged surface groups in aqueous solutions. III. Regulation
of the point of zero charge, surface dissociation constants and concentration of charged
surface groups of SiO2 by variation of the solution temperature or by modification with
sodium ions, J. Catal. 109: 41–50.
Alexiou, M. S., and Sermon, P. A. 1993. Aspects of the preparation of heterogeneous
catalysts by impregnation, React. Kinet. Catal. Lett. 51: 1–7.
Anderson, J. R. 1975. Structure of Metallic Catalysts, London: Academic Press.
Arakawa, H., Takeuchi, K., Matsuzaki, T., and Sugi, Y. 1984. Effect of metal dispersion
on the activity and selectivity of Rh/SiO2 catalyst for high pressure CO hydrogenation,
Chem. Lett. 1607–1610.
Arbabi, S., and Sahimi, M. 1991a. Computer simulations of catalyst deactivation – I. Model
formulation and validation, Chem. Eng. Sci. 46: 1739–1747.
Arbabi, S., and Sahimi, M. 1991b. Computer simulations of catalyst deactivation – II. The
effect of morphological transport and kinetic parameters on the performance of the
catalyst, Chem. Eng. Sci. 46: 1749–1755.
Ardiles, D. R. 1986. Activity and selectivity of nonuniform bifunctional catalysts. Analysis
of the fixed-bed reactor performance, Collect. Czech. Chem. Commun. 51: 2509–2520.
Ardiles, D. R., Scelza, O. A., and Castro, A. A. 1985. Activity and selectivity of non-uniform
bifunctional catalysts, Collect. Czech. Chem. Commun. 50: 726–737.
201
202 References
Ardiles, D. R., De Miguel, S. R., Castro, A. A., and Scelza, O. A. 1986. Pt–Re catalysts:
Study of the impregnation step, Appl. Catal. 24: 175–186.
Aris, R. 1975. The Mathematical Theory of Diffusion and Reaction in Permeable Catalysts,
Oxford: Clarendon Press.
Asua, J. M., and Delmon, B. 1987. Theoretical study of the influence of nonuniform active-
phase distribution on activity and selectivity of hydrodesulfurization catalysts, Ind. Eng.
Chem. Res. 26: 32–39.
Au, S. S., Dranoff, J. S., and Butt, J. B. 1995. Nonuniform activity distribution in catalyst
particles: Benzene hydrogenation on supported nickel in a single pellet diffusion reactor,
Chem. Eng. Sci. 50: 3801–3812.
Bacaros, T., Bebelis, S., Pavlou, S., and Vayenas, C. G. 1987. Optimal catalyst distribution
in pellets with shell progressive poisoning: The case of linear kinetics. In Catalyst Deac-
tivation 1987, eds. Delmon, B., and Froment, G. F., pp. 459–468. Amsterdam: Elsevier.
Baiker, A., and Holstein, W. L. 1983. Impregnation of alumina with copper chloride –
modeling of impregnation kinetics and internal copper profiles, J. Catal. 84: 178–188.
Baiker, A., Monti, D., and Wokaun, A. 1986. Impregnation of alumina with copper chloride:
Evidence for differently immobilized copper species, Appl. Catal. 23: 425–436.
Baratti, R., Cao, G., Morbidelli, M., and Varma, A. 1990. Optimal activity distribution in
nonuniformly impregnated catalyst particles: Numerical analysis, Chem. Eng. Sci. 45:
1643–1646.
Baratti, R., Wu, H., Morbidelli, M., and Varma, A. 1993. Optimal catalyst activity profiles
in pellets – X. The role of catalyst loading, Chem. Eng. Sci. 48: 1869–1881.
Baratti, R., Gavriilidis, A., Morbidelli, M., and Varma, A. 1994. Optimization of a non-
isothermal nonadiabatic fixed-bed reactor using Dirac-type silver catalysts for ethylene
epoxidation, Chem. Eng. Sci. 49: 1925–1936.
Baratti, R., Feckova, V., Morbidelli, M., and Varma, A. 1997. Optimal catalyst activity
profiles in pellets. 11. The case of multiple-step distributions, Ind. Eng. Chem. Res. 36:
3416–3420.
Barbier, J., and Marecot, P. 1981. Comparative study of structure sensitive reactions on
Pt/Al2 O3 and Ir/Al2 O3 catalysts, Nouveau J. Chim. 5: 393–396.
Barrow, N. J., and Bowden, J. W. 1987. A comparison of models for describing the adsorption
of anions on a variable charge mineral surface, J. Coll. Interf. Sci. 119: 236–250.
Bartholomew, C. H. 1993. Sintering kinetics of supported metals: New perspectives from a
unifying GPLE treatment, Appl. Catal. A 107: 1–57.
Bartholomew, C. H., and Farrauto, R. J. 1976. Chemistry of nickel–alumina catalysts,
J. Catal. 45: 41–53.
Bartholomew, C. H., Pannell, R. B., and Butler, J. L. 1980. Support and crystallite size
effects in CO hydrogenation on nickel, J. Catal. 65: 335–347.
Barto, M., Markos, J., and Brunovska, A. 1991. Oscillatory behaviour of a catalyst pellet
with narrow region of activity, Chem. Eng. Sci. 46: 2875–2880.
Basset, J. M., Dalmai-Imelik, G., Primet, M., and Mutin, R. 1975. A study of ben-
zene hydrogenation and identification of the adsorbed species with Pt/Al2 O3 catalysts,
J. Catal. 37: 22–36.
Beck, D. D., Sommers, J. W., and DiMaggio C. L. 1997. Axial characterization of catalytic
activity in close-coupled lightoff and underfloor catalytic converters, Appl. Catal. B 11:
257–272.
Becker, E. R., and Nuttall, T. A. 1979. Controlled catalyst distribution on supports by co-
impregnation. In Preparation of Catalysts II, eds. Delmon, B., Grange, P., Jacobs, P., and
Poncelet, G., pp. 159–169. Amsterdam: Elsevier.
Becker, E. R., and Wei, J. 1977a. Nonuniform distribution of catalysts on supports. I.
Bimolecular Langmuir reactions, J. Catal. 46: 365–371.
Becker, E. R., and Wei, J. 1977b. Nonuniform distribution of catalysts on supports. II. First
order reactions with poisoning, J. Catal. 46: 372–381.
References 203
Beeckman, J. W., and Hegedus, L. L. 1991. Design of monolith catalysts for power plant
NOx emission control, Ind. Eng. Chem. Res. 30: 969–978.
Benesi, H. A., Curtis, R. M., and Studer, H. P. 1968. Preparation of highly dispersed catalytic
metals. Platinum supported on silica gel, J. Catal. 10: 328–335.
Benjamin, M. M., and Leckie, J. O. 1981. Multiple-site adsorption of Cd, Cu, Zn and Pb on
amorphous iron oxyhydroxide, J. Coll. Interf. Sci. 79: 209–221.
Berenblyum, A. S., Mund, S. L., Karelskii, V. V., Goranskaya, T. P., Zolotukhin, V. E., and
Lakhman, L. I. 1986. Catalysts for the selective hydrogenation of unsaturated compounds
consisting of alumina particles with a regular palladium distribution, Kinet. and Catal.
27: 184–187.
Berger, D., and Pei, D. C. T. 1973. Drying of hygroscopic capillary porous solids – a theo-
retical approach, Int. J. Heat Mass Transfer 16: 293–302.
Blachou, V. M., and Philippopoulos, C. J. 1993. Adsorption of hexachloroplatinic acid on γ -
alumina coatings for preparation of monolithic structure catalysts, Chem. Eng. Commun.
119: 41–53.
Blanco, M. N., Caceres, C. V., Fierro, J. L. G., and Thomas, H. J. 1987. Influence of the
operative conditions on the preparation of pelleted Mo/Al2 O3 catalysts, Appl. Catal. 33:
231–244.
Bolan, N. S., and Barrow, N. J. 1984. Modelling the effect of adsorption of phosphate
and other anions on the surface charge of variable charge oxides, J. Soil Sci. 35: 273–
281.
Bonivardi, A. L., and Baltanas, M. A. 1990. Preparation of Pd/SiO2 catalysts for methanol
synthesis, J. Catal. 125: 243–259.
Borchert, A., and Buchholz, K. 1979. Inhomogeneous distribution of fixed enzymes inside
carriers, Biotechnol. Lett. 1: 15–20.
Borchert, A., and Buchholz, K. 1984. Improved biocatalyst effectiveness by controlled
immobilization of enzymes, Biotechnol. and Bioeng. 26: 727–736.
Bothe, N. 1982. Struktur, Stabilität und katalytische Aktivität sulfonsaurer Styrol-
Divinylbenzol-Harze, PhD Thesis, Technische Universität zu Braunschweig.
Bournonville, J. P., Franck, J. P., and Martino, G. 1983. Influence of the various activation
steps on the dispersion and the catalytic properties of platinum supported on chlorinated
alumina. In Preparation of Catalysts III, eds. Poncelet, G., Grange P., and Jacobs, P. A.,
pp. 81–90. Amsterdam: Elsevier.
Bowden, J. W., Posner, A. M., and Quirk, J. P. 1977. Ionic adsorption on variable charge
mineral surfaces. Theoretical charge development and titration curves, Aust. J. Soil. Res.
15: 121–136.
Bowden, J. W., Nagarajah, S., Barrow, N. J., Posner, A. M., and Quirk, J. P. 1980. Describing
the adsorption of phosphate, citrate and selenite on a variable-charge mineral surface,
Aust. J. Soil. Res. 18: 49–60.
Brunelle, J. P. 1978. Preparation of catalysts by adsorption of metal complexes on mineral
oxides, Pure Appl. Chem. 50: 1211–1229.
Brunelle, J. P., Sugler, A., and Le Page, J.-F. 1976. Active centers of platinum silica catalysts
in hydrogenolysis and isomerization of n-pentane, J. Catal. 43: 273–291.
Brunovska, A. 1987. Dynamic behaviour of a catalyst pellet with nonuniform activity dis-
tribution, Chem. Eng. Sci. 42: 1969–1976.
Brunovska, A. 1988. Dynamic behaviour of catalyst pellet with step function activity dis-
tribution, Chem. Eng. Sci. 43: 2546–2548.
Brunovska, A., Morbidelli, M., and Brunovsky, P. 1990. Optimal catalyst pellet activity
distributions for deactivating systems, Chem. Eng. Sci. 45: 917–925.
Brunovska, A., Remiarova, B., and Pranda, P. 1994. Role of catalyst pellet activity distri-
bution in catalyst poisoning, Appl. Catal. A 108: 141–156.
Buchholz, K. 1979. Nonuniform enzyme distribution inside carriers, Biotechnol. Lett. 1:
451–456.
204 References
Chen, H.-C., Gillies, G. C., and Anderson, R. B. 1980. Impregnating chromium and copper
in alumina, J. Catal. 62: 367–373.
Cheng, W. C., and Pereira, C. J. 1987. Preparation of cobalt–molybdenum hydrotreating
catalysts using hydrogen peroxide stabilization, Appl. Catal. 33: 331–341.
Chiang, C.-L., and Fang, Z.-R. 1994. Simulation and optimal design for the resid-
ual oil hydrodemetallation in a cocurrent moving-bed reactor, Chem. Eng. Sci. 49:
1175–1183.
Chiang, C.-L., and Tiou, H.-H. 1992. Optimal design for the residual oil hydrodemetallation
in a fixed bed reactor, Chem. Eng. Commun. 117: 383–399.
Chibata, I., Tosa, T., and Shibatani, T. 1992. The industrial production of optically active
compounds by immobolized biocatalysts. In Chirality in Industry, eds. Collins, A. N.,
Sheldrake, G. N., and Crosby, J., pp. 352–370. Wiley.
Chu, P., Petersen, E. E., and Radke, C. J. 1989. Modeling wet impregnation of nickel on γ -
alumina, J. Catal. 117: 52–70.
Chung, B. H., and Chang, H. N. 1986. Effect of internal diffusion on the apparent stability
of nonuniformly distributed biocatalysts, Korean J. Chem. Eng. 3: 39–43.
Clause, O., Kermarec, M., Bonneviot, L., Villain, F., and Che, M. 1992. Nickel(II) ion-
support interactions as a function of preparation method of silica-supported nickel ma-
terials, J. Am. Chem. Soc. 114: 4709–4717.
Contescu, C., and Vass, M. I. 1987. The effect of pH on the adsorption of palladium(II)
complexes on alumina, Appl. Catal. 33: 259–271.
Contescu, C., Hu, J., and Schwarz, J. A. 1993a. 1-pK multisites description of charge devel-
opment at the aqueous alumina interface, J. Chem. Soc. Faraday Trans. 89: 4091–4099.
Contescu, C., Jagiello, J., and Schwarz, J. A. 1993b. Heterogeneity of proton binding sites
at the oxide/solution interface, Langmuir 9: 1754–1765.
Contescu, C., Macovei, D., Craiu, C., Teodorescu, C., and Schwarz, J. A. 1995a. Thermal
induced evolution of chlorine-containing precursors in impregnated Pd/Al2 O3 catalysts,
Langmuir 11: 2031–2040.
Contescu, C., Jagiello, J., and Schwarz, J. A. 1995b. Proton affinity distributions: A scientific
basis for the design and construction of supported metal catalysts. In Preparation of
Catalysts VI, eds. Poncelet, G., Martens, J., Delmon, B., Jacobs, P. A., and Grange, P., pp.
237–252. Amsterdam: Elsevier.
Corbett, W. E., and Luss, D. 1974. The influence of non-uniform catalytic activity on the
performance of a single spherical pellet, Chem. Eng. Sci. 29: 1473–1483.
Cot, L., Guizard, C., and Larbot, A. 1988. Novel ceramic material for liquid separation
process: Present and prospective applications in microfiltration and ultrafiltration, Ind.
Ceram. 8: 143–148.
Coulson, J. M., and Richardson, J. F. 1991. Drying. In Chemical Engineering, 4th ed., vol.
2, Chapter 16. Oxford: Pergamon.
Cukierman, A. L., Laborde, M. A., and Lemcoff, N. O. 1983. Optimum activity distribution
in a catalyst pellet for a complex reaction, Chem. Eng. Sci. 38: 1977–1982.
Cusumano, J. A. 1991. Creating the future of the chemical industry – catalysts by molecular
design. In Perspectives in Catalysis, eds. Thomas, J. M., and Zamaraev, K. I., pp. 1–33.
Oxford: Blackwell Scientific.
Dadyburjor, D. B. 1982. Distribution for maximum activity of a composite catalyst,
A.I.Ch.E.J. 28: 720–728.
Dadyburjor, D. B. 1985. Selectivity over unifunctional multicomponent catalysts with
nonuniform distribution of components, Ind. Eng. Chem. Fundam. 24: 16–27.
Dadyburjor, D. B., and White, C. W., III 1990. Effect of position-dependent deactivation
on the design of a composite cracking catalyst, Chem. Eng. Sci. 45: 2619–2624.
D’Aniello, M. J. 1981. Anion adsorption on alumina, J. Catal. 69: 9–17.
Davis, J. A., and Leckie, J. O. 1978. Surface ionization and complexation at the oxide/water
interface 2. Surface properties of amorphous iron oxyhydroxide and adsorption of metal
ions, J. Coll. Interf. Sci. 67: 90–107.
206 References
Davis, J. A., and Leckie, J. O. 1980. Surface ionization and complexation at the oxide/water
interface 3. Adsorption of anions, J. Coll. Interf. Sci. 74: 32–43.
Davis, J. A., James, R. O., and Leckie J. O. 1978. Surface ionization and complexation at
the oxide/water interface 1. Computation of electrical double layer properties in simple
electrolytes, J. Coll. Interf. Sci. 63: 480–499.
De Jong, K. P. 1991. Deposition precipitation onto pre-shaped carrier bodies. Possibilities
and limitations. In Preparation of Catalysts V, eds. Poncelet, G., Jacobs, P. A., Grange, P.,
and Delmon, B., pp. 19–36. Amsterdam: Elsevier.
DeLancey, G. B. 1973. An optimal catalyst activation policy for poisoning problems, Chem.
Eng. Sci. 28: 105–118.
Delannay, F. (ed.) 1984. Characterization of Heterogeneous Catalysts. New York: Marcel
Dekker.
Delmon, B., and Houalla, M. 1979. Tentative classification of the factors influencing the
reduction step in the activation of supported catalysts. In Preparation of Catalysts II, eds.
Delmon, B., Grange, P., Jacobs, P., and Poncelet, G., pp. 439–468. Amsterdam: Elsevier.
De Miguel, S. R., Scelza, O. A., Castro, A. A., Baronetti, G. T., Ardiles, D. R., and Parera,
J. M. 1984. Radial profiles in Pt/Al2 O3 , Re/Al2 O3 , and Pt–Re/Al2 O3 , Appl. Catal. 9:
309–315.
D’Espinose de la Caillerie, J. -B., Bobin, C., Rebours, B., and Clause, O. 1995. Alumina/water
interfacial phenomena during impregnation. In Preparation of Catalysts VI, eds. Poncelet,
G., Martens, J., Delmon, B., Jacobs, P., and Grange, P., pp. 169–184. Amsterdam: Elsevier.
Do, D. D. 1984. A method for solving diffusion and reaction problems with nonuniform
activity catalysts, Chem. Eng. Sci. 39: 1519–1522.
Do, D. D. 1985. Modeling of impregnation kinetics and internal activity profiles: Adsorption
of HCl, HReO4 and H2 PtCl6 onto γ -Al2 O3 . Chem. Eng. Sci. 40: 1871–1880.
Do, D. D., and Bailey, J. E. 1982. Approximate analytical solutions for porous catalysts with
nonuniform activity, Chem. Eng. Sci. 37: 545–551.
Dorling, T. A., Lynch, B. W. J., and Moss, R. L. 1971. The structure and activity of supported
metal catalysts. V. Variables in the preparation of platinum/silica catalysts, J. Catal. 20:
190–201.
Dorling, T. A., and Moss, R. L. 1967. The structure and activity of supported metal catalysts.
II. Crystallite size and CO chemisorption in platinum/silica catalysts, J. Catal. 7: 378–385.
Dougherty, R. C., and Verykios, X. E. 1987. Nonuniformly activated catalysts, Catal. Rev. –
Sci. Eng. 29: 101–150.
Downing, D. M., Lee, J. W., and Butt, J. B. 1979. Simulation models for intraparticle deac-
tivation: Scope and reliability, A.I.Ch.E.J. 25: 461–469.
Duncombe, P. R., and Weller, S. W. 1985. Thermal behavior of shell molybdena–alumina
catalysts, A.I.Ch.E.J. 31: 410–414.
Engels, S., Lausch, H., and Schwokowski, R. 1987. Untersuchungen zur Adsorption bi- und
polyfunktioneller organischer Saeuren an Aluminiumoxid, Chem. Techn. (Leipzig) 39:
387–391.
Ernst, W. R., and Daugherty, D. J. 1978. Method for the study of performance of a single
spherical particle with nonuniform catalytic activity, A.I.Ch.E.J. 24: 935–937.
Fenelonov, V. B. 1975. Analysis of the drying stages in the technology of deposited catalysts
I. Role of the pore structure and the drying conditions, Kinet. and Catal. 16: 628–635.
Fenelonov, V. B., Neimark, A. V., and Kheifets, L. I. 1979. Analysis of steps of impreg-
nation and drying in preparation of supported catalysts. In Preparation of Catalysts
II, eds. Delmon, B., Grange, P., Jacobs, P., and Poncelet, G., pp. 233–244. Amsterdam:
Elsevier.
Finlayson, B. A. 1980. Nonlinear Analysis in Chemical Engineering. New York: McGraw-
Hill.
Foger, K. 1984. Dispersed metal catalysts. In Catalysis Science and Technology, vol. 6,
eds. Anderson, J. R., and Boudart, M., pp. 227–305. Berlin: Springer-Verlag.
References 207
Frost, A. C., Sawyer, J. E., Summers, J. C., Shah, Y. T., and Dassori, C. G. 1990. Reac-
tor model for the fixed bed catalytic incineration of VOC’s, presented at the 83rd An-
nual Meeting and Exhibition, Air and Waste Management Association, Pittsburgh, PA,
June 24–29.
Fuentes, S., and Figueras, F. 1980. The influence of particle size on the catalytic properties
of alumina-supported rhodium catalysts, J. Catal. 61: 443–453.
Fujiyama, T., Otsuka, M., Tsuiki, H., and Ueno, A. 1987. Control of the impregnation
profile of Ni in an Al2 O3 sphere, J. Catal. 104: 323–330.
Furusaki, S. 1988. Engineering aspects of immobilised biocatalysts, J. Chem. Eng. Japan 21:
219–230.
Galiasso, R., de Ochoa, O. L., and Andreu, P. 1983. Influence of the drying rate on the
distribution of active metals and on the activity of heavy crude hydrotreating catalyst,
Appl. Catal. 5: 309–322.
Gates, G. C., Winouskas, J. S., and Heath, H. W. 1972. The dehydration of t-butyl alcohol
catalyzed by sulfonic acid resin, J. Catal. 24: 320–327.
Gavriilidis, A., and Varma, A. 1992. Optimal catalyst activity profiles in pellets: 9. Study of
ethylene epoxidation, A.I.Ch.E.J. 38: 291–296.
Gavriilidis, A., Varma, A., and Morbidelli, M. 1993a. Optimal distribution of catalyst in
pellets, Catal. Rev. – Sci. Eng. 35: 399–456.
Gavriilidis, A., Sinno, B., and Varma, A. 1993b. Influence of loading on metal surface area
for Ag/α-Al2 O3 catalysts, J. Catal. 139: 41–47.
Gazimzyanov, N. R., Mikhailov, V. I., and Zadko, I. I. 1992. Study of distribution of active
components in alumino–nickel–molybdenum catalysts obtained by a single impregna-
tion, Kinet. and Catal. 33: 737–743.
Gelosa, D., Pedeferri, M. P., and Morbidelli, M. 1998. Esterification reactions with nonuni-
formly sulfonated resins, First Annual Report INTREASEP Project, Joule Contract
JOE3-CT97-0082.
Gladden, L. F. 1995. Applications of nuclear magnetic resonance imaging in particle tech-
nology, Part. Part. Syst. Charact. 12: 59–67.
Gottifredi, J. C., Gonzo, E. E., and Quiroga, O. D. 1981. Isothermal effectiveness factor II.
Analytical expression for single reaction with arbitrary kinetics, geometry and activity
distribution, Chem. Eng. Sci. 36: 713–719.
Goula, M. A., Kordulis, C., Lycourghiotis, A., and Fierro, J. L. G. 1992a. Development of
molybdena catalysts supported on γ -alumina extrudates with four different Mo profiles:
Preparation, characterisation and catalytic properties, J. Catal. 137: 285–305.
Goula, M. A., Kordulis, C., and Lycourghiotis, A. 1992b. Influence of impregnation param-
eters on the axial Mo/γ -alumina profiles studied using a novel simple technique, J. Catal.
133: 486–497.
Guisan, J. M., Serrano, J., Melo, F. V., and Ballesteros, A. 1987. Mixed enzymic reaction–
internal diffusion kinetics of nonuniformly distributed immobilised enzymes. The system
agarose–micrococcal endonuclease, Appl. Biochem. Biotech. 14: 49–72.
Hachiya, K., Sasaki, M., Saruta, Y., Mikami, N., and Yasunaga, T. 1984. Static and kinetic
studies of adsorption–desorption of metal ions on a γ -Al2 O3 surface. 1. Static study of
adsorption–desorption, J. Phys. Chem. 88: 23–27.
Hamada, H. Funaki, R., Kuwahara, Y., Kintaichi, Y., Wakabayashi, K., and Ito, T. 1987.
Systematic preparation of supported Rh catalysts having desired metal particle size by
using silica supports with controlled pore structure, Appl. Catal. 30: 177–180.
Hanika, J., and Ehlova, V. 1989. Modelling of internal diffusion inside a catalyst particle with
non-uniform radial activity profile for parallel reactions, Collect. Czech. Chem. Commun.
54: 81–90.
Hanika, J., Machek, V., Nemec, V., Ruzicka, V., and Kunz, J. 1982. Simultaneous diffusion
and adsorption of chloroplatinic acid in charcoal pellet during preparation process of
supported platinum catalyst, J. Catal. 77: 248–256.
208 References
Hanika, J., Hao, L. H., and Ruzicka, V. 1983. Relation between the conditions of preparation
and the activity of supported platinum catalysts, Collect. Czech. Chem. Commun. 48:
3079–3085.
Hanika, J., Janousek, V., and Sporka, K. 1987a. Modelling of impregnation of γ -alumina
with cobalt and molybdenum salts. CoCl2 –(NH4 )2 MoO4 –γ -Al2 O3 aluminate type sys-
tem, Collect. Czech. Chem. Commun. 52: 663–671.
Hanika, J., Janousek, V., and Sporka, K. 1987b. Modelling of impregnation of γ -alumina
with cobalt and molybdenum salts. Co(NO3 )2 –(NH4 )2 MoO4 –γ -Al2 O3 chloride type sys-
tem, Collect. Czech. Chem. Commun. 52: 672–677.
Harold, M. P., and Luss, D. 1987. Impact of the catalytic activity profile on observed multi-
plicity features: CO oxidation on Pt/Al2 O3 , Ind. Eng. Chem. Res. 26: 1616–1621.
Harriott, P. 1969. Diffusion effects in the preparation of impregnated catalysts, J. Catal. 14:
43–48.
Haworth, A. 1990. A review of the modelling of sorption from aqueous solution, Adv. Coll.
Interf. Sci. 32: 43–78.
Hayes, K. F., and Leckie, J. O. 1986. Mechanism of lead ion adsorption at the goethite–water
interface, ACS Symp. Ser. 323: 114–141.
Heck, R. M., and Farrauto, R. J. 1995. Catalytic Air Pollution Control. Commercial Tech-
nology, New York: Van Nostrand Reinhold.
Hegedus, L. L. 1980. Catalyst pore structures by constrained nonlinear optimisation, Ind.
Eng. Chem. Prod. Res. Dev. 19: 533–537.
Hegedus, L. L., and Gumbleton, J. J. 1980. Catalysts, computers and cars: A growing sym-
biosis, Chemtech, October, 630–642.
Hegedus, L. L., and McCabe, R. 1984. Catalyst Poisoning. New York: Marcel Dekker.
Hegedus, L. L., and Pereira, C. J. 1990. Reaction engineering for catalyst design, Chem.
Eng. Sci. 45: 2027–2044.
Hegedus, L. L., and Summers, J. C. 1977. Improving the poison resistance of supported
catalysts, J. Catal. 48: 345–353.
Hegedus, L. L., Summers, J. C., Schlatter, J. C., and Baron, K. 1979a. Poison resistant
catalysts for the simultaneous control of hydrocarbon, carbon monoxide and nitrogen
oxide emissions, J. Catal. 56: 321–335.
Hegedus, L. L., Chou, T. S., Summers, J. C., and Potter, N. M. 1979b. Multicomponent
chromatographic processes during the impregnation of alumina pellets with noble metals.
In Preparation of Catalysts II, eds. Delmon, B., Grange, P., Jacobs, P., and Poncelet, G.,
pp. 171–184. Amsterdam: Elsevier.
Heise, M. S., and Schwarz, J. A. 1985. Preparation of metal distributions within catalyst
supports I. Effect of pH on catalytic metal profiles, J. Coll. Interf. Sci. 107: 237–243.
Heise, M. S., and Schwarz, J. A. 1986. Preparation of metal distributions within catalyst
supports II. Effect of ionic strength on catalytic metal profiles, J. Coll. Interf. Sci. 113:
55–61.
Heise, M. S., and Schwarz, J. A. 1987. Preparation of metal distributions within catalyst
supports. In Preparation of Catalysts IV, eds. Delmon, B., Grange, P., Jacobs, P. A., and
Poncelet, G., pp. 1–13. Amsterdam: Elsevier.
Heise, M. S., and Schwarz, J. A. 1988. Preparation of metal distributions within catalyst
supports III. Single component modeling of pH, ionic strength and concentration effects,
J. Coll. Interf. Sci. 123: 51–58.
Hepburn, J. S., Stenger, H. G., and Lyman, C. E. 1989a. Distributions of HF co-impregnated
rhodium, platinum and palladium in alumina honeycomb supports, Appl. Catal. 55: 271–
285.
Hepburn, J. S., Stenger, H. G., and Lyman, C. E. 1989b. Effects of drying on the preparation
of HF co-impregnated rhodium/Al2 O3 catalysts, Appl. Catal. 55: 287–299.
Hepburn, J. S., Stenger, H. G., and Lyman, C. E. 1989c. Co-impregnation of rhodium into
alumina honeycombs with acids and salts, Appl. Catal. 56: 107–118.
References 209
Jianguo, W., Jiayu, Z., and Li, P. 1983. The role of competitive adsorbate in the impregnation
of platinum in pelleted alumina support. In Preparation of Catalysts III, eds. Poncelet,
G., Grange P., and Jacobs, P. A., pp. 57–67. Amsterdam: Elsevier.
Jiratova, K. 1981. Isoelectric point of modified alumina, Appl. Catal. 1: 165–167.
Johnson, D. L., and Verykios, X. E. 1983. Selectivity enhancement in ethylene oxidation
employing partially impregnated catalysts, J. Catal. 79: 156–163.
Johnson, D. L., and Verykios, X. E. 1984. Effects of radially nonuniform distributions of
catalytic activity on performance of spherical catalyst pellets, A.I.Ch.E.J. 30: 44–50.
Juang, H.-D., and Weng, H.-S. 1983. Performance of catalysts with nonuniform activity
profiles 2. Theoretical analysis for nonisothermal reactions. Ind. Eng. Chem. Fundam.
22: 224–230.
Juang, H.-D., and Weng, H.-S. 1984. Performance of biocatalysts with nonuniformly dis-
tributed immobilized enzyme, Biotechnol. and Bioeng. 26, 623–626.
Juang, H.-D., Weng, H.-S. and Wang, C.-C. 1981. The performance of catalysts with nonuni-
form activity profile I. Theoretical analysis for isothermal reactions, Stud. Surf. Sci. Catal.
7: 866–876.
Karakonstantis, L., Kordulis, Ch., and Lycourghiotis, A. 1992. Mechanism of adsorption
of tungstates on the interface of γ -alumina/electrolyte solutions, Langmuir 8: 1318–
1324.
Karpe, P., and Ruckenstein, E. 1989. Role of electrokinetic phenomena in the preparation
of supported metal catalysts, Coll. and Polym. Sci. 267: 145–150.
Kasaoka, S., and Sakata, Y. 1968. Effectiveness factors for non-uniform catalyst pellets,
J. Chem. Eng. Japan 1: 138–142.
Kehoe, P. 1974. Letter to the Editors, Chem. Eng. Sci. 29: 315.
Kehoe, J. P. G., and Butt, J. B. 1972. Interactions of inter- and intraphase gradients in a
diffusion limited catalytic reaction, A.I.Ch.E.J. 18: 347–355.
Keil, F. J., and Rieckmann, C. 1994. Optimisation of three-dimensional catalyst pore struc-
tures, Chem. Eng. Sci. 49: 4811–4822.
Keller, J. B., Falkovitz, M. S., and Frisch, H. 1984. Optimal catalyst distribution in a mem-
brane, Chem. Eng. Sci. 39: 601–604.
Kempling, J. C., and Anderson, R. B. 1970. Hydrogenolysis of n-butane on supported
ruthenium, Ind. Eng. Chem. Proc. Des. Dev. 9: 116–120.
Kheifets, L. I., Neimark, A. V., and Fenelonov, V. B. 1979. Analysis of the influence of
permeation and drying stages in supported catalyst preparation on the distribution of
components I. Precipitation of one component from solution, Kinet. and Catal. 20: 626–
632.
Kim, D. S., Kurusu, Y., Wachs, I. E., Hardcastle, F. D., and Segawa, K. 1989. Physicochemical
properties of MoO3 –TiO2 prepared by an equilibrium adsorption method, J. Catal. 120:
325–336.
Kimura, H. 1993. Selective oxidation of glycerol on a platinum–bismuth catalyst by using
a fixed bed reactor, Appl. Catal. A 105: 147–158.
King, D. L. 1978. A Fischer–Tropsch study of supported ruthenium catalysts, J. Catal. 51:
386–397.
Klein, J., Widdecke, H., and Bothe, N. (1984). Influence of functional group distribution
on the thermal stability and catalytic activity of sulfonated styrene-DVB-copolymers,
Macromol. Chem. and Phys., Suppl. 6: 211–226.
Klugherz, P. D., and Harriott, P. 1971. Kinetics of ethylene oxidation on a supported silver
catalyst, A.I.Ch.E.J. 17: 856–866.
Ko, S. H., and Chou, T. C. 1994. Hydrogenation of (–)-α-pinene over nickel–
phosphorus/aluminum oxide catalysts prepared by electroless deposition, Can. J. Chem.
Eng. 72: 862–873.
Komiyama, M. 1985. Design and preparation of impregnated catalysts, Catal. Rev. – Sci.
Eng. 27: 341–372.
References 211
Komiyama, M., and Muraki, H. 1990. Design of intrapellet activity profiles in supported
catalysts. In Handbook of Heat and Mass Transfer, vol. 4, ed. Cheremisinoff, N. P.,
pp. 447–500. Houston: Gulf Publishing Company.
Komiyama, M., Ohashi, K., Morioka, Y., and Kobayashi, J. 1997. Effects of the intrapellet
activity profiles on the selectivities in consecutive reactions: 1,3-butadiene hydrogenation
over Pt/Al2 O3 , Bull. Chem. Soc. Japan 70: 1009–1013.
Komiyama, M., Merrill, R. P., and Harnsberger, H. F. 1980. Concentration profiles in im-
pregnation of porous catalysts: Nickel on alumina, J. Catal. 63: 35–52.
Kotter, M., and Riekert, L. 1979. The influence of impregnation, drying and activation on
the activity and distribution of CuO on α-alumina. In Preparation of Catalysts II, eds.
Delmon, B., Grange, P., Jacobs, P., and Poncelet, G., pp. 51–63. Amsterdam: Elsevier.
Kotter, M., and Riekert, L. 1983a. Impregnation-type catalysts with nonuniform distribu-
tion of the active component. Part I: Influence of the accessibility of the active component
on activity and selectivity, Chem. Eng. Fundam. 2: 22–30.
Kotter, M., and Riekert, L. 1983b. Impregnation-type catalysts with nonuniform distri-
bution of the active component. Part II: Preparation and properties of catalysts with
different distribution of the active component on inert carriers, Chem. Eng. Fundam. 2:
31–38.
Kresge, C. T., Chester, A. W., and Oleck, S. M. 1992. Control of metal radial profiles in
alumina supports by carbon dioxide, Appl. Catal. A 81: 215–226.
Kulkarni, S. S., Mauze. G. R., and Schwarz, J. A. 1981. Concentration profiles and the design
of metal-supported catalysts, J. Catal. 69: 445–453.
Kummer, J. T. 1980. Catalysts for automobile emission control, Prog. Energy Combust. Sci.
6: 177–199.
Kummert, R., and Stumm, W. 1980. The surface complexation of organic acids on hydrous
γ -Al2 O3 , J. Colloid Interface Sci. 75: 373–385.
Kunimori, K., Nakajima, I., and Uchijima, T. 1982. Catalytic performance of egg white type
Pt/Al2 O3 catalyst in the oxidation of C3 H8 and CO, Chem. Lett. 1165–1168.
Kunimori, K., Kawasaki, E., Nakajima, I., and Uchijima, T. 1986. Catalytic performance of
egg-white type Pt/Al2 O3 catalyst: Multiple steady states in the oxidation of CO, Appl.
Catal. 22: 115–122.
Lee, C. K., and Varma, A. 1987. Steady state multiplicity behavior of an adiabatic plug-flow
reactor with nonuniformly active catalyst, Chem. Eng. Commun. 58: 287–309.
Lee, C. K., and Varma, A. 1988. An isothermal fixed-bed reactor with nonuniformly active
catalysts: Experiments and theory, Chem. Eng. Sci. 43: 1995–2000.
Lee, C. K., Morbidelli, M., and Varma, A. 1987a. Optimal catalyst activity profiles in pellets
6. Optimization of the isothermal fixed-bed reactor with multiple zones, Ind. Eng. Chem.
Res. 26: 167–170.
Lee, C. K., Morbidelli, M., and Varma, A. 1987b. Steady state multiplicity behavior of an
isothermal axial dispersion fixed-bed reactor with nonuniformly active catalyst, Chem.
Eng. Sci. 42: 1595–1608.
Lee, H. H. 1981. Generalized effectiveness factor for pellets with nonuniform activity dis-
tribution: Asymptotic region of strong diffusion effect, Chem. Eng. Sci. 36: 1921–1925.
Lee, H. H. 1984. Catalyst preparation by impregnation and activity distribution, Chem.
Eng. Sci. 39: 859–864.
Lee, H. H. 1985. Catalyst preparation by impregnation and activity distributions – II. Coim-
pregnation for nonuniform distribution based on competitive adsorption, Chem. Eng.
Sci. 40: 1295–1300.
Lee, H. H., and Ruckenstein, E. 1983. Catalyst sintering and reactor design, Cat. Rev. – Sci.
Eng. 25: 475–550.
Lee, J. W., Butt, J. B., and Downing, D. M. 1978. Kinetic, transport and deactivation rate in-
teractions on steady state and transient responses in heterogeneous catalysis, A.I.Ch.E.J.
24: 212–222.
212 References
Martin, G. R., White, C. W., III, and Dadyburjor, D. B. 1987. Design of zeolite/silica–alumina
catalysts for triangular cracking reactions, J. Catal. 106: 116–124.
Masi, M., Sangalli, M., Carra, S., Cao, G., and Morbidelli, M. 1988. Kinetics of ethylene
hydrogenation on supported platinum. Analysis of multiplicity and nonuniformly active
catalyst particle behavior, Chem. Eng. Sci. 43: 1849–1854.
Mathur, I., Bakhshi, N. N., and Mathews, J. F. 1972. Effect of ultrasonic waves on the activity
of chromia–alumina catalyst, Can. J. Chem. Eng. 50: 344–348.
McCoy, M. 1999. Catalyst makers look for growth, Chemical and Engineering News,
September 1999, pp. 17–25.
Melis, S., Varma, A., and Pereira, C. J. 1997. Optimal distribution of catalyst for a case
involving heterogeneous and homogeneous reactions, Chem. Eng. Sci. 52: 165–169.
Melo, F., Cervello, J., and Hermana, E. 1980a. Impregnation of porous supports – I. The-
oretical study of the impregnation of one or two active species, Chem. Eng. Sci. 35:
2165–2174.
Melo, F., Cervello, J., and Hermana, E. 1980b. Impregnation of porous supports – II. System
Ni/Ba on alumina, Chem. Eng. Sci. 35: 2175–2184.
Michalko, E. 1966a. Method for oxidizing gaseous combustible waste products, U.S. Patent
3,259,454.
Michalko, E. 1966b. Preparation of catalyst for the treatment of combustible waste products,
U.S. Patent 3,259,589.
Mieth, J. A., Schwarz, J. A., Huang, Y.-J., and Fung, S. C. 1990. The effect of chloride on
the point of zero charge of γ -Al2 O3 , J. Catal. 122: 202–205.
Minhas, S., and Carberry, J. J. 1969. On the merits of partially impregnated catalysts,
J. Catal. 14: 270–272.
Morbidelli, M., and Varma, A. 1982. Optimal catalyst activity profiles in pellets 2. The
influence of external mass transfer resistance, Ind. Eng. Chem. Fundam. 21: 284–289.
Morbidelli, M., and Varma, A. 1983. On shape normalization for non-uniformly active
catalyst pellets II, Chem. Eng. Sci. 38: 297–305.
Morbidelli, M., Servida, A., and Varma, A. 1982. Optimal catalyst activity profiles in pellets
1. The case of negligible external mass transfer resistance, Ind. Eng. Chem. Fundam. 21:
278–284.
Morbidelli, M., Servida, A., Paludetto, R., and Carra, S. 1984a. Optimal catalyst design for
ethylene oxide synthesis, J. Catal. 87: 116–125.
Morbidelli, M., Servida, A., and Varma, A. 1984b. Optimal distribution of immobilized
enzyme in a pellet for a substrate-inhibited reaction, Biotechnol. and Bioeng. 26: 1508–
1510.
Morbidelli, M., Servida, A., Carra, S., and Varma, A., 1985. Optimal catalyst activity profiles
in pellets 3. The nonisothermal case with negligible external transport limitations, Ind.
Eng. Chem. Fundam. 24: 116–119.
Morbidelli, M., Servida, A., and Varma, A. 1986a. Optimal catalyst activity profiles in
pellets. 4. Analytical evaluation of the isothermal fixed-bed reactor, Ind. Eng. Chem.
Fundam. 25: 307–313.
Morbidelli, M., Servida, A., Carra, S., and Varma, A. 1986b. Optimal catalyst activity profiles
in pellets. 5. Optimization of the isothermal fixed-bed reactor, Ind. Eng. Chem. Fundam.
25: 313–321.
Morbidelli, M., Brunovska, A., Wu, H., and Varma, A. 1991. Authors’ reply to comments
by S. Pavlou et al., Chem. Eng. Sci. 46: 3328–3329.
Moss, R. L., Pope, D., Davis, B. J., and Edwards, D. H. 1979. The structure and activity of
supported metal catalysts. VIII. Chemisorption and benzene hydrogenation on palla-
dium/silica catalysts, J. Catal. 58: 206–219.
Moyers, C. G., and Baldwin, G. W. 1997. Psychrometry, evaporative cooling and solids
drying. In Perry’s Chemical Engineers’ Handbook, 7th ed., eds. Perry, R. H., Green,
D. W., and Maloney, J. O., Chapter 12. New York: McGraw-Hill.
214 References
Mulcahy, F. M., Houalla, M., and Hercules, D. M. 1987. The effect of the isoelectric point
on the adsorption of molybdates on fluoride modified aluminas, J. Catal. 106: 210–215.
Mulcahy, F. M., Fay, M. J., Proctor, A., Houalla, M., and Hercules, D. M. 1990. The adsorption
of metal oxyanions on alumina, J. Catal. 124: 231–240.
Naoki, B., Katsuyuki, O., and Shigeki, S. 1996. Numerical approach for improving the
conversion characteristics of exhaust catalysts under warming-up condition, SAE Paper
962076.
Narsimhan, G. 1976. Catalyst dilution as a means to establish an optimum temperature
profile, Ind. Eng. Chem. Proc. Des. Dev. 15: 302–307.
Neimark, A. V., Fenelonov, V. B., and Heifets, L. I. 1976. Analysis of the drying stage in
the technology of supported catalysts, React. Kinet. Catal. Lett. 5: 67–72.
Neimark, A. V., Kheifez, L. I., and Fenelonov, V. B. 1981. Theory of preparation of supported
catalysts. Ind. Eng. Chem. Prod. Res. Dev. 20: 439–450.
Nicolescu, I. V., Parvulescu, V., Parvulescu, V., and Angelescu, E. 1990. Preparation of the
metallic supported catalysts from precursors with different immobilization capacities on
γ -Al2 O3 surface, Rev. Roum. Chim. 35: 145–159.
Nyström, M. 1978. Effectiveness factors for non-uniform catalytic activity of a spherical
pellet, Chem. Eng. Sci. 33: 379–382.
Ochoa, O., Galiasso, R., and Andreu, P. 1979. Study of some variables involved in the
preparation of impregnated catalysts for the hydrotreating of heavy oils. In Preparation
of Catalysts II, eds. Delmon, B., Grange, P., Jacobs, P., and Poncelet, G., pp. 493–506.
Amsterdam: Elsevier.
Oh, S. H., and Cavendish, J. C. 1982. Transients of monolithic catalytic converters: Response
to step changes in feedstream temperature as related to controlling automobile emissions,
Ind. Eng. Chem. Prod. Res. Dev. 21: 29–37.
Oh, S. H., and Cavendish, J. C. 1983. Design aspects of poison-resistant automobile mono-
lithic catalysts, Ind. Eng. Chem. Prod. Res. Dev. 22: 509–518.
Ollis, D. F. 1972. Diffusion infuences in denaturable insolubilized enzyme catalysts, Biotech-
nol. and Bioeng. 24: 871–884.
Olsbye, U., Wendelbo, R., and Akporiaye, D. 1997. Study of Pt/alumina catalysts prepara-
tion, Appl. Catal. A 152: 127–141.
Otsuka, M., Fujiyama, T., Tsukamoto, Y., Tsuiki, H., and Ueno, A. 1987. Control of
the impregnation profile of Co in an Al2 O3 sphere, Bull. Chem. Soc. Japan 60: 2881–
2885.
Ott, R. J., and Baiker, A. 1983. Impregnation of γ -alumina with copper chloride. Equi-
librium behaviour, impregnation profiles and immobilization kinetics. In Preparation of
Catalysts III, eds. Poncelet, G., Grange P., and Jacobs, P. A., pp. 685–696. Amsterdam:
Elsevier.
Papa, J., and Shah, Y. T. 1992. An approximate solution for effectiveness factor for a square
wave shell catalyst, Chem. Eng. Commun. 113: 55–61.
Papadakis, V. G., Pliangos, C. A., Yentekakis, I. V., Verykios, X. E., and Vayenas, C. G.,
1996. Development of high performance, Pd-based, three way catalysts, Catal. Today 29:
71–75.
Papageorgiou, P. 1984. Preparation of Pt/γ -Al2 O3 pellets with internal step-distribution of
catalyst, MS Thesis, University of Notre Dame.
Papageorgiou, P., Price, D. M., Gavriilidis, A., and Varma, A. 1996. Preparation of Pt/γ -
Al2 O3 pellets with internal step-distribution of catalyst: Experiments and theory, J. Catal.
158: 439–451.
Parfitt, G. D. 1976. Surface chemistry of oxides, Pure Appl. Chem. 48: 415–418.
Park, S. H., Lee, S. B., and Ryu, D. D. Y. 1981. Design of a nonuniformly distributed
biocatalyst, Biotechnol. and Bioeng. 23: 2591–2600.
Parks, G. A. 1965. The isoelectric points of solid oxides, solid hydroxides, and aqueous
hydroxo complex systems, Chem. Rev. 65: 177–198.
References 215
Pavlou, S., and Vayenas, C. G. 1990a. Optimal catalyst activity distribution in pellets for
selectivity maximization in triangular nonisothermal reaction systems: Application to
cases of light olefin epoxidation, J. Catal. 122: 389–405.
Pavlou, S., and Vayenas, C. G. 1990b. Optimal catalyst activity profile in pellets with shell-
progressive poisoning: The case of fast linear kinetics, Chem. Eng. Sci. 45: 695–703.
Pavlou, S., Vayenas, C. G., and Dassios. G. 1991. Comments on optimal catalyst activity
profiles in pellets. VIII. General nonisothermal reacting systems with arbitrary kinetics,
Chem. Eng. Sci. 46: 3327–3328.
Pereira, C. J., Kubsh, J. E., and Hegedus, L. L. 1988. Computer-aided design of catalytic
monoliths for automobile emission control, Chem. Eng. Sci. 43: 2087–2094.
Pernicone, N., and Traina, F. 1984. Commercial catalyst preparation, Appl. Ind. Catal. 3:
1–24.
Pfefferle, L. D., and Pfefferle, W. C. 1987. Catalysis in combustion, Catal. Rev. – Sci. Eng.
29: 219–267.
Pirkle, J. C., and Wachs, I. E. 1987. Activity profiling in catalytic reactors, Chem. Eng. Prog.,
August, 29–34.
Podolski, W. F., and Kim, Y. G. 1974. Modeling the water-gas shift reaction, Ind. Eng. Chem.
Proc. Des. Dev. 13: 415–421.
Polomski, R. E., and Wolf, E.E. 1978. Deactivation of a composite automobile catalyst
pellet II. Bimolecular Langmuir reaction, J. Catal. 52: 272–279.
Post, M. F. M., and Sie, S. T. 1986. Process for the preparation of hydrocarbons, Eur. Patent
Appl. 0174696.
Pranda, P., and Brunovska, A. 1993. Optimal catalyst pellet activity distributions for deac-
tivating systems: Experimental study, Chem. Eng. Sci. 48: 3423–3430.
Price, D. M., and Varma, A. 1988. Effective diffusivity measurement through an adsorbing
porous solid, AIChE Symp. Ser. 266: 88–96
Psyllos, A., and Philippopoulos, C. 1993. Performance of a monolithic catalytic converter
used in automotive emission control: The effect of longitudinal parabolic active metal
distribution, Ind. Eng. Chem. Res. 32: 1555–1559.
Radovich, J. M. 1985. Mass transfer effects in fermentations using immobilised whole cells,
Enzyme Microb. Technol. 7: 2–10.
Ray, W. H. 1981. Advanced Process Control, New York: McGraw-Hill.
Remiarova, B., Brunovska, A., and Morbidelli, M. 1993. Sulphur poisoning of nonuniformly
impregnated Pt–alumina catalysts, Chem. Eng. Sci. 48: 1227–1236.
Reuel, R. C., and Bartholomew, C. H. 1984. Effects of support and dispersion on the CO
hydrogenation activity/selectivity properties of cobalt, J. Catal. 85: 78–88.
Righetto, L., Azimonti, G., Missana, T., and Bidoglio, G. 1995. The triple layer model
revised, Coll. and Surf. A 95: 141–157.
Roth, J. F., and Reichard, T. E. 1972. Determination and effect of platinum concentration
profiles in supported catalysts, J. Res. Inst. Catal. Hokkaido Univ. 20: 85–94.
Ruckenstein, E. 1970. The effectiveness of diluted porous catalysts, A.I.Ch.E.J. 16: 151–153.
Ruckenstein, E., and Dadyburjor, D. B. 1983. Sintering and redispersion in supported metal
catalysts, Rev. Chem. Eng. 1: 251–354.
Ruckenstein, E., and Karpe, P. 1989. Control of metal distribution in supported catalysts
by pH, ionic strength and coimpregnation, Langmuir 5: 1393–1407.
Ruckenstein, E., and Malhotra, M. L. 1976. Splitting of platinum crystallites supported on
thin, nonporous alumina films, J. Catal. 41: 303–311.
Ruthven, D. M. 1984. Principles of Adsorption and Adsorption Processes, New York: Wiley.
Rutkin, D. R., and Petersen, E. E. 1979. The effect on selectivity of the macroscopic distri-
bution of the components in a dual function catalyst, Chem. Eng. Sci. 34: 109–116.
Sadhukhan, P., and Petersen, E. E. 1976. Oxidation of napthalene in packed-bed reactor
with catalyst activity profile: A design scheme for improved reactor stability and higher
product yield, A.I.Ch.E.J. 22: 808–810.
216 References
Santacesaria, E., Carra, S., and Adami, I. 1977a. Adsorption of hexachloroplatinic acid on
γ -alumina, Ind. Eng. Chem. Prod. Res. Dev. 16: 41–44.
Santacesaria, E., Gelosa, D., and Carra, S. 1977b. Basic behavior of alumina in the presence
of strong acids, Ind. Eng. Chem. Prod. Res. Dev. 16: 45–46.
Santacesaria, E., Galli, C., and Carra, S. 1977c. Kinetic aspects of the impregnation of
alumina pellets with hexachloroplatinic acid, React. Kinet. Catal. Lett. 6: 301–306.
Santhanam, N., Conforti, T. A., Spieker, W., and Regalbuto, J. R. 1994. Nature of metal
catalyst precursors adsorbed onto oxide supports, Catal. Today 21: 141–156.
Saracco, G., and Speccia, V. 1994. Catalytic inorganic membrane reactors: Present experi-
ence and future opportunities, Catal. Rev. Sci. Eng. 36: 305–384.
Sarkany, J., and Gonzalez, R. D. 1983. Effect of pretreatment on dispersion and structure of
silica- and alumina-supported Pt catalysts, Ind. Eng. Chem. Prod. Res. Dev. 22: 548–552.
Satterfield, C. N., Pelossof, A. A., and Sherwood, T. K. 1969. Mass transfer limitations in a
trickle-bed reactor, A.I.Ch.E.J. 15: 226–234.
Scelza, O. A., Castro, A. A., Ardiles, D. R., and Parera, J. M. 1986. Modeling of the im-
pregnation step to prepare supported Pt/Al2 O3 catalysts. Ind. Eng. Chem. Fundam. 25:
84–88.
Schbib, N. S., Garcia, M. A., Gigola, C. E., and Errazu, A. F. 1996. Kinetics of front-end
acetylene hydrogenation in ethylene production, Ind. Eng. Chem. Res. 35: 1496–1505.
Scheuch, S., Kamphuis, A. J., McKay, I. R., and Walls, J. R. 1996. A two stage flu-
idized/spouted bed for the granulation of catalyst powder. In Proceedings of the 1996
IChemE Research Event, pp. 997–999.
Schilling, L. B. 1994. Catalysis and Biocatalysis Technologies. Leveraging Resources and
Targeting Performance, National Institute of Standards and Technology.
Scholten, J. J. F., and Van Montfoort, A. 1962. The determination of the free-metal surface
area of palladium catalysts, J. Catal. 1: 85–92.
Schwarz, J. A. 1992. The adsorption/impregnation of catalytic precursors on pure and com-
posite oxides, Catal. Today 15: 395–405.
Schwarz, J. A., and Heise, M. S. 1990. Preparation of metal distributions within catalyst
supports IV. Multicomponent effects, J. Coll. Interf. Sci. 135: 461–467.
Schwarz, J. A., Ugbor, C. T., and Zhang, R. 1992. The adsorption/impregnation of Pd(II)
cations on alumina, silica and their composite oxides, J. Catal. 138: 38–54.
Schwarz, J. A., Contescu, C., and Contescu, A. 1995. Methods for preparation of catalytic
materials, Chem. Rev. 95: 477–510.
Seyedmonir, S. R., Plischke, J. K., Vannice, M. A., and Young, H. W. 1990. Ethylene oxida-
tion over small silver crystallites, J. Catal. 123: 534–549.
Shadman-Yazdi, F., and Petersen, E. E. 1972. Changing catalyst performance by varying
the distribution of active catalyst within porous supports, Chem. Eng. Sci. 27: 227–237.
Shah, A. M., and Regalbuto, J. R. 1994. Retardation of Pt adsorption over oxide supports
at pH extremes: Oxide dissolution or high ionic strength? Langmuir 10: 500–504.
Sheintuch, M., Lev, O., Mendelbaum, S., and David, B. 1986. Optimal feed distribution in
reactions with maximal rate, Ind. Eng. Chem. Fundam. 25: 228–233.
Shelley, S. 1997. Destroying emissions with catalysts, Chem. Eng., July, pp. 57–60.
Sherrington, D. C., and Hodge, P. 1988. Syntheses and Separations Using Functional Poly-
mers, New York: Wiley.
Shyr, Y.-S., and Ernst, W. R. 1980. Preparation of nonuniformly active catalysts, J. Catal.
63: 425–432.
Sivaraj, Ch., Contescu, Cr., and Schwarz, J. A. 1991. Effect of calcination temperature of
alumina on the adsorption/impregnation of Pd(II) compounds, J. Catal. 132: 422–431.
Smirniotis, P. G., and Ruckenstein, E. 1993. The activity of composite catalysts: Theory and
experiments for spherical and cylindrical pellets, Chem. Eng. Sci. 48: 585–593.
Smith, T. G., and Carberry, J. J. 1975. On the use of partially impregnated catalysts for yield
enhancement in non-isothermal non-adiabatic fixed bed reactors, Can. J. Chem. Eng. 53:
347–349.
References 217
Soled, S. L., Baumgartner, J. E., Reyes, S. C., and Iglesia, E. 1995. Synthesis of eggshell
cobalt catalysts by molten salt impregnation techniques. In Preparation of Catalysts VI,
eds. Poncelet, G., Martens, J., Delmon, B., Jacobs, P. A., and Grange, P., pp. 989–997.
Amsterdam: Elsevier.
Souza, G. L. M., Santos, A. C. B., Lovate, D. A., and Faro, A. C. 1989. Characterization of
Ni-Mo/Al2 O3 catalysts prepared by simultaneous impregnation of the active compo-
nents, Catal. Today 5: 451–461.
Spanos, N., Vordonis, L., Kordulis, C., and Lycourghiotis, A. 1990a. Molybdenum–oxo
species deposited on alumina by adsorption. I. Mechanism of the adsorption, J. Catal.
124: 301–314.
Spanos, N., Vordonis, L., Kordulis, C., Koutsoukos, P. G., and Lycourghiotis, A. 1990b.
Molybdenum–oxo species deposited on alumina by adsorption. II. Regulation of the
surface MoVI concentration by control of the protonated surface hydroxyls, J. Catal. 124:
315–323.
Spanos, N., Kordulis, C., and Lycourghiotis, A. 1991. Development of a methodology for
investigating the adsorption of species containing catalytically active ions on the surface
of industrial carriers. In Preparation of Catalysts V, eds. Poncelet, G., Jacobs, P. A., Grange,
P., and Delmon, B., pp. 175–184. Amsterdam: Elsevier.
Spanos, N., Matralis, H. K., Kordulis, C., and Lycourghiotis, A. 1992. Molybdenum–oxo
species deposited on titania by adsorption: Mechanism of the adsorption and character-
ization of the calcined species, J. Catal. 136: 432–445.
Spielbauer, D., Zeilinger, H., and Knoezinger, H. 1993. Adsorption of palladium–ammino–
aquo complexes on γ -alumina and silica, Langmuir 9: 460–466.
Sporka, K., and Hanika, J. 1992. Development of cobalt–molybdenum hydrodesulfurization
catalysts, Collect. Czech. Chem. Commun. 57: 2501–2508.
Srinivasan, R., Liu, H. C., and Weller, S. W. 1979. Sintering of shell molybdena–alumina
catalysts, J. Catal. 57: 87–95.
Stenger, H. G., Meyer, E. C., Hepburn, J. S., and Lyman, C. E. 1988. Nitric oxide reduction
using co-impregnated rhodium on an alumina Celcor honeycomb, Chem. Eng. Sci. 43:
2067–2072.
Stern, O. 1924. Zur Theorie der elektrolytischen Doppelschicht, Z. Elektrochemie 30: 508–
516.
Stiles, A. B., and Koch, T. A. 1995. Catalyst Manufacture, 2nd ed., New York: Marcel Dekker.
Subramanian, S., and Schwarz, J. A. 1991. Adsorption of chloroplatinic acid and chloroiridic
acid on composite oxides, Langmuir 7: 1436–1440.
Subramanian, S., Obrigkeit, D. D., Peters, C. R., and Chattha, M. S. 1992. Adsorption of
palladium on γ -alumina, J. Catal. 138: 400–404.
Summers, J. C., and Ausen, S. A. 1978. Catalyst impregnation: Reactions of noble metal
complexes with alumina, J. Catal. 52: 445–452.
Summers, J. C., and Hegedus, L. L. 1978. Effects of platinum and palladium impregnation
on the performance and durability of automobile exhaust oxidizing catalysts, J. Catal.
51: 185–192.
Summers, J. C., and Hegedus, L. L. 1979. Modes of catalyst deactivation in stoichiometric
automobile exhaust, Ind. Eng. Chem. Prod. Res. Dev. 18: 318–324.
Szegner, J. 1997. Effects of nonuniform catalyst distribution on inorganic membrane reactor
performance: Experiments and theory, PhD Thesis, University of Notre Dame.
Szegner, J., Yeung, K. L., and Varma, A. 1997. Effect of catalyst distribution in a membrane
reactor: Experiments and model, A.I.Ch.E.J. 43: 2059–2072.
Tanabe, K. 1970. Solid Acids and Bases, New York: Academic Press.
Taylor, K. C. 1984. Automobile catalytic converters. In Catalysis Science and Technol-
ogy, vol. 5, eds. Anderson, J. R., and Boudart, M., pp. 119–170. Berlin: Springer-
Verlag.
Thayer, A. M. 1994. Catalyst industry stresses need for partners as key to future success,
Chem. and Eng. News, July 11, pp. 19–20.
218 References
Tosa, T., Mori, T., Fuse, N., and Chibata, I. 1969. Studies on continuous enzyme reactions.
Part V. Kinetics and industrial application of aminoacylase column for continuous optical
resolution of acyl-DL-amino acids, Agr. Biol. Chem. 33: 1047–1052.
Tronci, S., Baratti, R., and Gavriilidis, A. 1999. Catalytic converter design for minimisation
of cold-start emissions, Chem. Eng. Commun., 173: 53–77.
Tsotsis, T. T., Champagnie, A. M., Minet, R. G., and Liu, P. K. T. 1993. Catalytic membrane
reactors. In Computer Aided Design of Catalysts, eds. Becker, E. R., and Pereira, C. J.,
pp. 471–552. New York: Marcel Dekker.
Uemura, Y., and Hatate, Y. 1989. Effect of nickel concentration profile on selectivity of
acetylene hydrogenation, J. Chem. Eng. Japan 22: 287–291.
Uemura, Y., Hatate, Y., and Ikari, A. 1986. Effects of post-impregnation drying conditions
on physical properties and overall reaction rate of nickel/alumina catalysts, J. Chem. Eng.
Japan 19: 560–566.
Uemura, Y., Hatate, Y., and Ikari, A. 1987. Formation of nickel concentration profile in
nickel/alumina catalyst during post-impregnation drying, J. Chem. Eng. Japan 20: 117–
123.
Underwood, R. P., and Bell, A. T. 1987. Influence of particle size on carbon monoxide
hydrogenation over silica and lanthana supported rhodium, Appl. Catal. 34: 289–310.
Uzio, D., Giroir-Fendler, A., Lieto, J., and Dalmon, J. A. 1991. Catalytic membrane reactors:
Preparation and characterization of an active Pt/alumina membrane, Key Eng. Mater. 61:
111–116.
Van den Berg, G. H., and Rijnten, H. Th. 1979. The impregnation and drying step in catalyst
manufacturing. In Preparation of Catalysts II, eds. Delmon, B., Grange, P., Jacobs, P., and
Poncelet, G., pp. 265–277, Elsevier, Amsterdam.
Van Erp, W. A., Nanne, J. M., and Post, M. F. M. 1986. Catalyst preparation, Eur. Patent
Appl. 0178008.
Van Herwijnen, T., Van Doesburg, H., and De Jong, W. A. 1973. Kinetics of the methanation
of CO and CO2 on a nickel catalyst, J. Catal. 28: 391–402.
Van Tiep, L., Bureau-Tardy, M., Bugli, G., Djega-Mariadassou, G., Che, M., and Bond, G.
C. 1986. The effect of reduction conditions on the chloride content of Ir/TiO2 catalysts
and their activity for benzene hydrogenation, J. Catal. 99: 449–460.
Varghese, P., and Wolf, E. E. 1980. Effectiveness and deactivation of a diluted catalyst
pellet, A.I.Ch.E.J. 26: 55–60.
Vayenas, C. G., and Pavlou, S. 1987a. Optimal catalyst activity distribution and generalized
effectiveness factors in pellets: Single reactions with arbitrary kinetics, Chem. Eng. Sci.
42: 2633–2645.
Vayenas, C. G., and Pavlou, S. 1987b. Optimal catalyst distribution for selectivity maximiza-
tion in pellets: Parallel and consecutive reactions, Chem. Eng. Sci. 42: 1655–1666.
Vayenas, C. G., and Pavlou, S. 1988. Optimal catalyst distribution for selectivity maxi-
mization in nonisothermal pellets: The case of parallel reactions, Chem. Eng. Sci. 43:
2729–2740.
Vayenas, C. G., and Verykios, X. E. 1989. Optimisation of catalytic activity distributions in
porous pellets. In Handbook of Heat and Mass Transfer, vol. 3, ed. Cheremisinoff, N. P.,
pp. 135–181. Houston: Gulf Publishing Company.
Vayenas, C. G., Pavlou, S., and Pappas, A. D. 1989. Optimal catalyst distribution for selec-
tivity maximization in nonisothermal pellets: The case of consecutive reactions, Chem.
Eng. Sci. 44: 133–145.
Vayenas, C. G., Verykios, X. E., Yentekakis, I. B., Papadakis, E. G., and Pliangos, C. A. 1994.
New three-way catalysts with Pt, Rh and Pd, each supported on a separate support, Eur.
Patent Appl. 0665047A1.
Vazquez, P. G., Caceres, C. V., Blanco, M. N., and Thomas, H. J. 1993. Influence of support
on concentration profiles in alumina pellets impregnated with molybdenum solutions,
Int. Commun. Heat Mass Transfer 20: 631–642.
References 219
Verykios, X. E., Kluck, R. W., and Johnson, D. L. 1983. Fixed-bed reactor simulation with
nonuniformly activated catalyst pellets. In Modeling and Simulation in Engineering, eds.
Ames, W. F., and Vichnevetsky, R., pp. 3–10. Amsterdam: North-Holland.
Vieth, W. R., Mendiratta, A. K., Mogensen, A. O., Saini, R., and Venkatasubramanian, K.
1973. Mass transfer and biochemical reaction in enzyme membrane reactor systems – I.
Single enzyme reactions, Chem. Eng. Sci. 28: 1013–1020.
Villadsen, J. 1976. The effectiveness factor for an isothermal pellet with decreasing activity
towards the pellet surface, Chem. Eng. Sci. 31: 1212–1213.
Villadsen, J., and Michelsen, M. L. 1978. Solution of Differential Equation Models by Poly-
nomial Approximation. Englewood Cliffs, NJ: Prentice-Hall.
Vincent, R. C., and Merrill, R. P. 1974. Concentration profiles in impregnation of porous
catalysts, J. Catal. 35: 206–217.
Voltz, S. E., Morgan, C. R., Liederman, D., and Jacob, S. M. 1973. Kinetic study of carbon
monoxide and propylene oxidation on platinum catalysts, Ind. Eng. Chem. Prod. Res.
Dev. 12: 294–301.
Vordonis, L., Koutsoukos, P. G., and Lycourghiotis, A. 1984. Regulation of the point of
zero charge and surface acidity constants of γ -Al2 O3 using sodium and fluoride ions as
modifiers, J. Chem. Soc. Chem. Commun. 1309–1310.
Vordonis, L., Koutsoukos, P. G., and Lycourghiotis, A. 1986a. Development of carriers with
controlled concentration of charged surface groups in aqueous solutions. I. Modification
of γ -Al2 O3 with various amounts of sodium ions, J. Catal. 98: 296–307.
Vordonis, L., Koutsoukos, P. G., and Lycourghiotis, A. 1986b. Development of carriers with
controlled concentration of charged surface groups in aqueous solutions. II. Modification
of γ -Al2 O3 with various amounts of lithium and fluoride ions, J. Catal. 101: 186–194.
Vordonis, L., Koutsoukos, P. G., and Lycourghiotis, A. 1990. Adsorption of molybdates on
doped γ -aluminas in alkaline solutions, Coll. and Surf. 50: 353–361.
Vordonis, L., Spanos, N., Koutsoukos, P. G., and Lycourghiotis, A. 1992. Mechanism of
adsorption of Co2+ and Ni2+ ions on the pure and fluorinated γ -alumina/electrolyte
solution interface, Langmuir 8: 1736–1743.
Wang, J. B., and Varma, A. 1978. Effectiveness factors for pellets with step-distribution of
catalyst, Chem. Eng. Sci. 33: 1549–1552.
Wang, L., and Hall, W. K. 1980. On the genesis of molybdena–alumina catalyst, J. Catal.
66: 251–255.
Wang, L., and Hall, W. K. 1982. The preparation and genesis of molybdena–alumina and
related catalyst systems, J. Catal. 77: 232–241.
Wang, T., and Schmidt, L. D. 1981. Intraparticle redispersion of Rh and Pt–Rh particles on
SiO2 and Al2 O3 by oxidation–reduction cycling, J. Catal. 70: 187–197.
Wanke, S. E., and Flynn, P. C. 1975. The sintering of supported metal catalysts, Catal. Rev. –
Sci. Eng. 12: 93–135.
Wei, J., and Becker, E. R. 1975. The optimum distribution of catalytic material on support
layers in automotive catalysis, Adv. Chem. Ser. 143: 116–132.
Westall, J. C. 1986. Reactions at the oxide–solution interface: Chemical and electrostatic
models, ACS Symp. Ser. 323: 54–78.
Westall, J., and Hohl, H. 1980. A comparison of electrostatic models for the oxide/solution
interface, Adv. Coll. Interf. Sci. 12: 265–294.
White, C. W., III, and Dadyburjor, D. B. 1989. Effect of nonuniform deactivation on the
optimum design of composite cracking catalysts, Chem. Eng. Commun. 86: 113–124.
Widdecke, H., Klein, J., and Haupt, U. 1986. The influence of matrix structure on the activity
and selectivity of polymer supported catalysts, Macromol. Chem., Macromol. Symp. 4:
145–155.
Wilson, G. R., and Hall, W. K. 1970. Studies of the hydrogen held by solids. XVIII. Hydrogen
and oxygen chemisorption on alumina- and zeolite-supported platinum, J. Catal. 17: 190–
206.
220 References
Wolf, E. E. 1977. Activity and lifetime of a composite automobile catalyst pellet, J. Catal.
47: 85–91.
Wu, C. H., and Hammerle, R. H. 1983. Development of low cost, thermally stable, mono-
lithic three-way catalyst system, Ind. Eng. Chem. Prod. Res. Dev. 22: 559–565.
Wu, H. 1994. Application of optimal catalyst activity distribution theory, A.I.Ch.E.J. 40:
2060–2064.
Wu, H. 1995. Optimal feed distribution in isothermal fixed-bed reactors for parallel reac-
tions, Chem. Eng. Sci. 50: 441–451.
Wu, H., Yuan, Q., and Zhu, B. 1988. An experimental investigation of optimal active catalyst
distribution in nonisothermal pellets, Ind. Eng. Chem. Res. 27: 1169–1174.
Wu, H., Brunovska, A., Morbidelli, M., and Varma, A. 1990a. Optimal catalyst activity
profiles in pellets VIII. General nonisothermal reacting systems with arbitrary kinetics,
Chem. Eng. Sci. 45: 1855–1862; 46: 3328–3329.
Wu, H., Yuan, Q., and Zhu, B. 1990b. An experimental study of optimal active catalyst
distribution in pellets for maximum selectivity, Ind. Eng. Chem. Res. 29: 1771–1776.
Xidong, H., Yongnian, Y., and Jiayu, Z. 1988. Influence of soluble aluminium on the state
and surface properties of platinum in a series of reduced platinum/alumina catalysts,
Appl. Catal. 40: 291–313.
Xue, J., Huang, Y.-J., and Schwarz, J. A. 1988. Interaction between iridium and platinum
precursors in the preparation of iridium–platinum catalysts, Appl. Catal. 42: 61–76.
Yates, D. E., Levine, S., and Healy, T. W. 1974. Site-binding model of the electrical double
layer at the oxide/water interface, J. Chem. Soc. Faraday Trans. I 70: 1807–1818.
Ye, J., and Yuan, Q. 1992. Optimal catalyst activity distribution for effectiveness factor,
productivity and selectivity maximization in generalized nonisothermal reacting systems
with arbitrary kinetics, Chem. Eng. Sci. 47: 615–621.
Yeung, K. L., Aravind, R., Zawada, R. J. X., Szegner, J., Cao, G., and Varma, A. 1994.
Nonuniform catalyst distribution for inorganic membrane reactors: Theoretical consid-
erations and preparation techniques, Chem. Eng. Sci. 49: 4823–4838.
Yeung, K. L., Aravind, R., Szegner, J., and Varma, A. 1996. Metal composite membranes:
Synthesis, characterisation and reaction studies, Stud. Surf. Sci. Catal., 101: 1349–1358.
Yeung, K. L., Sebastian, J. M., and Varma, A. 1997. Mesoporous alumina membranes:
Synthesis, characterization, thermal stability and nonuniform distribution of catalyst, J.
Membr. Sci. 131: 9–28.
Yeung, K. L., Gavriilidis, A., Varma, A., and Bhasin, M. M. 1998. Effects of 1,2-
dichloroethane addition on the optimal silver catalyst distribution in pellets for epoxi-
dation of ethylene, J. Catal. 174: 1–12.
Yortsos, Y. C., and Tsotsis, T. T. 1981. On the relationship between the effectiveness factor
for the Robin and Dirichlet problem for a catalyst with variable catalytic activity, Chem.
Eng. Sci. 36: 1734–1736.
Yortsos, Y. C., and Tsotsis, T. T. 1982a. Asymptotic behavior of the effectiveness factor for
variable activity catalysts, Chem. Eng. Sci. 37: 237–243.
Yortsos, Y. C., and Tsotsis, T. T. 1982b. On the relationship between the effectiveness
factors for the Robin and Dirichlet problems for a catalyst with nonuniform catalytic
activity. The case of generalized isothermal and nonisothermal kinetics, Chem. Eng. Sci.
37: 1436–1437.
Zaki, M. I., Mansour, S. A. A., Taha, F., and Mekhemer, G. A. H. 1992. Chromia on
silica and alumina catalysts: Surface structural consequences of interfacial events in the
impregnation course of aquated Cr(III) ions, Langmuir 8: 727–732.
Zaman, J., and Chakma, A. 1994. Inorganic membrane reactors, J. Membr. Sci., 92: 1–28.
Zhang, R., and Schwarz, J. A. 1992. Design of inhomogeneous metal distributions within
catalyst particles, Appl. Catal. A. 91: 57–65.
Zhukov, A. N. 1996. Dependences of the point of zero charge and the isoelectric point
of amphoteric solid surface on concentration and degree of binding of the ions of the
background electrolyte. The case of 1 : 1 electrolyte, Coll. J. (Engl. Ed.) 58: 270–272.
Author Index
221
222 Author Index
225
226 Subject Index